]> git.saurik.com Git - bison.git/blame - NEWS
gnulib: update
[bison.git] / NEWS
CommitLineData
ed7658fe 1GNU Bison NEWS
3af4feb2 2
7ca7a3ab
AD
3* Noteworthy changes in release ?.? (????-??-??) [?]
4
476c1cca
AD
5** Bug fixes
6
7*** Location of errors
8
9 In C++ parsers, out-of-bounds errors can happen when a rule with an empty
10 ride-hand side raises a syntax error. The behavior of the default parser
11 (yacc.c) in such a condition was undefined.
12
13 Now all the parsers match the behavior of glr.c: @$ is used as the
14 location of the error. This handles gracefully rules with and without
15 rhs.
7ca7a3ab 16
c4e686a1 17* Noteworthy changes in release 3.0.4 (2015-01-23) [stable]
11d82017 18
3e82dfaa
AD
19** Bug fixes
20
21*** C++ with Variants (lalr1.cc)
22
23 Fix a compiler warning when no %destructor use $$.
11d82017 24
658b1893
AD
25*** Test suites
26
27 Several portability issues in tests were fixed.
28
da6f0746 29* Noteworthy changes in release 3.0.3 (2015-01-15) [stable]
0bd5ee5f 30
461f1516
AD
31** Bug fixes
32
ee028dce
AD
33*** C++ with Variants (lalr1.cc)
34
35 Problems with %destructor and '%define parse.assert' have been fixed.
36
37*** Named %union support (yacc.c, glr.c)
827bc59c
AD
38
39 Bison 3.0 introduced a regression on named %union such as
40
41 %union foo { int ival; };
42
ee028dce 43 The possibility to use a name was introduced "for Yacc compatibility".
827bc59c
AD
44 It is however not required by POSIX Yacc, and its usefulness is not clear.
45
ee028dce 46*** %define api.value.type union with %defines (yacc.c, glr.c)
8d4dc896 47
ee028dce
AD
48 The C parsers were broken when %defines was used together with "%define
49 api.value.type union".
8d4dc896 50
461f1516
AD
51*** Redeclarations are reported in proper order
52
53 On
54
55 %token FOO "foo"
56 %printer {} "foo"
57 %printer {} FOO
58
59 bison used to report:
60
61 /tmp/foo.yy:2.10-11: error: %printer redeclaration for FOO
62 %printer {} "foo"
63 ^^
64 /tmp/foo.yy:3.10-11: previous declaration
65 %printer {} FOO
66 ^^
67
68 Now, the "previous" declaration is always the first one.
69
0bd5ee5f 70
b167e7ba
AD
71** Documentation
72
73 Bison now installs various files in its docdir (which defaults to
74 '/usr/local/share/doc/bison'), including the three fully blown examples
75 extracted from the documentation:
76
77 - rpcalc
78 Reverse polish calculator, a simple introductory example.
79 - mfcalc
80 Multi-function Calc, a calculator with memory and functions and located
81 error messages.
82 - calc++
83 a calculator in C++ using variant support and token constructors.
84
a3efd13e 85* Noteworthy changes in release 3.0.2 (2013-12-05) [stable]
6d94eebb 86
7b0ca050
AD
87** Bug fixes
88
184b42c8
AD
89*** Generated source files when errors are reported
90
91 When warnings are issued and -Werror is set, bison would still generate
92 the source files (*.c, *.h...). As a consequence, some runs of "make"
93 could fail the first time, but not the second (as the files were generated
94 anyway).
95
96 This is fixed: bison no longer generates this source files, but, of
97 course, still produces the various reports (*.output, *.xml, etc.).
98
be29c71d
AD
99*** %empty is used in reports
100
101 Empty right-hand sides are denoted by '%empty' in all the reports (text,
102 dot, XML and formats derived from it).
103
7b0ca050
AD
104*** YYERROR and variants
105
106 When C++ variant support is enabled, an error triggered via YYERROR, but
107 not caught via error recovery, resulted in a double deletion.
6d94eebb 108
c2ecada3 109* Noteworthy changes in release 3.0.1 (2013-11-12) [stable]
e386b50f 110
0a244a22
AD
111** Bug fixes
112
39bace5d
AD
113*** Errors in caret diagnostics
114
115 On some platforms, some errors could result in endless diagnostics.
e4678430
AD
116
117*** Fixes of the -Werror option
118
119 Options such as "-Werror -Wno-error=foo" were still turning "foo"
120 diagnostics into errors instead of warnings. This is fixed.
121
122 Actually, for consistency with GCC, "-Wno-error=foo -Werror" now also
123 leaves "foo" diagnostics as warnings. Similarly, with "-Werror=foo
124 -Wno-error", "foo" diagnostics are now errors.
e386b50f 125
071863b3
AD
126*** GLR Predicates
127
128 As demonstrated in the documentation, one can now leave spaces between
129 "%?" and its "{".
130
265640d5
AD
131*** Installation
132
133 The yacc.1 man page is no longer installed if --disable-yacc was
134 specified.
135
39bace5d
AD
136*** Fixes in the test suite
137
138 Bugs and portability issues.
139
534497f5 140* Noteworthy changes in release 3.0 (2013-07-25) [stable]
facb910c 141
8458a411 142** WARNING: Future backward-incompatibilities!
597afd73 143
597afd73
AD
144 Like other GNU packages, Bison will start using some of the C99 features
145 for its own code, especially the definition of variables after statements.
146 The generated C parsers still aim at C90.
147
8458a411 148** Backward incompatible changes
47db7ed1
AD
149
150*** Obsolete features
151
40bb6f78
AD
152 Support for YYFAIL is removed (deprecated in Bison 2.4.2): use YYERROR.
153
154 Support for yystype and yyltype is removed (deprecated in Bison 1.875):
155 use YYSTYPE and YYLTYPE.
156
157 Support for YYLEX_PARAM and YYPARSE_PARAM is removed (deprecated in Bison
158 1.875): use %lex-param, %parse-param, or %param.
47db7ed1 159
05e25f23
AD
160 Missing semicolons at the end of actions are no longer added (as announced
161 in the release 2.5).
162
1fa19a76
AD
163*** Use of YACC='bison -y'
164
165 TL;DR: With Autoconf <= 2.69, pass -Wno-yacc to (AM_)YFLAGS if you use
166 Bison extensions.
167
168 Traditional Yacc generates 'y.tab.c' whatever the name of the input file.
169 Therefore Makefiles written for Yacc expect 'y.tab.c' (and possibly
170 'y.tab.h' and 'y.outout') to be generated from 'foo.y'.
171
172 To this end, for ages, AC_PROG_YACC, Autoconf's macro to look for an
173 implementation of Yacc, was using Bison as 'bison -y'. While it does
174 ensure compatible output file names, it also enables warnings for
175 incompatibilities with POSIX Yacc. In other words, 'bison -y' triggers
176 warnings for Bison extensions.
177
178 Autoconf 2.70+ fixes this incompatibility by using YACC='bison -o y.tab.c'
179 (which also generates 'y.tab.h' and 'y.output' when needed).
180 Alternatively, disable Yacc warnings by passing '-Wno-yacc' to your Yacc
181 flags (YFLAGS, or AM_YFLAGS with Automake).
182
597afd73
AD
183** Bug fixes
184
c21e515e 185*** The epilogue is no longer affected by internal #defines (glr.c)
597afd73
AD
186
187 The glr.c skeleton uses defines such as #define yylval (yystackp->yyval) in
188 generated code. These weren't properly undefined before the inclusion of
189 the user epilogue, so functions such as the following were butchered by the
190 preprocessor expansion:
191
192 int yylex (YYSTYPE *yylval);
193
6c7022f7 194 This is fixed: yylval, yynerrs, yychar, and yylloc are now valid
597afd73
AD
195 identifiers for user-provided variables.
196
f0f95a50
AD
197*** stdio.h is no longer needed when locations are enabled (yacc.c)
198
199 Changes in Bison 2.7 introduced a dependency on FILE and fprintf when
200 locations are enabled. This is fixed.
201
bb4b189b
AD
202*** Warnings about useless %pure-parser/%define api.pure are restored
203
597afd73
AD
204** Diagnostics reported by Bison
205
206 Most of these features were contributed by Théophile Ranquet and Victor
207 Santet.
73370a9d 208
016426c1
TR
209*** Carets
210
211 Version 2.7 introduced caret errors, for a prettier output. These are now
212 activated by default. The old format can still be used by invoking Bison
213 with -fno-caret (or -fnone).
214
fec5f3c0
AD
215 Some error messages that reproduced excerpts of the grammar are now using
216 the caret information only. For instance on:
217
218 %%
219 exp: 'a' | 'a';
220
221 Bison 2.7 reports:
222
223 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
224 in.y:2.12-14: warning: rule useless in parser due to conflicts: exp: 'a' [-Wother]
225
226 Now bison reports:
227
228 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
229 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
230 exp: 'a' | 'a';
231 ^^^
232
233 and "bison -fno-caret" reports:
234
235 in.y: warning: 1 reduce/reduce conflict [-Wconflicts-rr]
236 in.y:2.12-14: warning: rule useless in parser due to conflicts [-Wother]
237
1048a1c9 238*** Enhancements of the -Werror option
518e8830 239
1048a1c9 240 The -Werror=CATEGORY option is now recognized, and will treat specified
d949eefd
AD
241 warnings as errors. The warnings need not have been explicitly activated
242 using the -W option, this is similar to what GCC 4.7 does.
1048a1c9
AD
243
244 For example, given the following command line, Bison will treat both
d949eefd 245 warnings related to POSIX Yacc incompatibilities and S/R conflicts as
1048a1c9
AD
246 errors (and only those):
247
248 $ bison -Werror=yacc,error=conflicts-sr input.y
249
250 If no categories are specified, -Werror will make all active warnings into
251 errors. For example, the following line does the same the previous example:
252
253 $ bison -Werror -Wnone -Wyacc -Wconflicts-sr input.y
254
255 (By default -Wconflicts-sr,conflicts-rr,deprecated,other is enabled.)
256
257 Note that the categories in this -Werror option may not be prefixed with
258 "no-". However, -Wno-error[=CATEGORY] is valid.
259
260 Note that -y enables -Werror=yacc. Therefore it is now possible to require
261 Yacc-like behavior (e.g., always generate y.tab.c), but to report
262 incompatibilities as warnings: "-y -Wno-error=yacc".
518e8830 263
46bdb8ec 264*** The display of warnings is now richer
73370a9d 265
46bdb8ec 266 The option that controls a given warning is now displayed:
73370a9d 267
46bdb8ec 268 foo.y:4.6: warning: type clash on default action: <foo> != <bar> [-Wother]
73370a9d 269
46bdb8ec
TR
270 In the case of warnings treated as errors, the prefix is changed from
271 "warning: " to "error: ", and the suffix is displayed, in a manner similar
d949eefd 272 to GCC, as [-Werror=CATEGORY].
1048a1c9 273
46bdb8ec
TR
274 For instance, where the previous version of Bison would report (and exit
275 with failure):
1048a1c9
AD
276
277 bison: warnings being treated as errors
46bdb8ec 278 input.y:1.1: warning: stray ',' treated as white space
1048a1c9 279
46bdb8ec 280 it now reports:
1048a1c9 281
1048a1c9
AD
282 input.y:1.1: error: stray ',' treated as white space [-Werror=other]
283
284*** Deprecated constructs
285
286 The new 'deprecated' warning category flags obsolete constructs whose
287 support will be discontinued. It is enabled by default. These warnings
288 used to be reported as 'other' warnings.
289
73370a9d 290*** Useless semantic types
9641b918
VS
291
292 Bison now warns about useless (uninhabited) semantic types. Since
293 semantic types are not declared to Bison (they are defined in the opaque
294 %union structure), it is %printer/%destructor directives about useless
295 types that trigger the warning:
296
297 %token <type1> term
298 %type <type2> nterm
299 %printer {} <type1> <type3>
300 %destructor {} <type2> <type4>
301 %%
302 nterm: term { $$ = $1; };
303
304 3.28-34: warning: type <type3> is used, but is not associated to any symbol
305 4.28-34: warning: type <type4> is used, but is not associated to any symbol
306
31557b9e 307*** Undefined but unused symbols
b921d92f 308
31557b9e
AD
309 Bison used to raise an error for undefined symbols that are not used in
310 the grammar. This is now only a warning.
b921d92f
VS
311
312 %printer {} symbol1
313 %destructor {} symbol2
31557b9e 314 %type <type> symbol3
b921d92f
VS
315 %%
316 exp: "a";
317
73370a9d 318*** Useless destructors or printers
ea9a35c6
VS
319
320 Bison now warns about useless destructors or printers. In the following
321 example, the printer for <type1>, and the destructor for <type2> are
322 useless: all symbols of <type1> (token1) already have a printer, and all
323 symbols of type <type2> (token2) already have a destructor.
324
325 %token <type1> token1
326 <type2> token2
327 <type3> token3
328 <type4> token4
329 %printer {} token1 <type1> <type3>
330 %destructor {} token2 <type2> <type4>
331
d87ea54c
AD
332*** Conflicts
333
334 The warnings and error messages about shift/reduce and reduce/reduce
335 conflicts have been normalized. For instance on the following foo.y file:
336
337 %glr-parser
338 %%
339 exp: exp '+' exp | '0' | '0';
340
341 compare the previous version of bison:
342
343 $ bison foo.y
344 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
345 $ bison -Werror foo.y
346 bison: warnings being treated as errors
347 foo.y: conflicts: 1 shift/reduce, 2 reduce/reduce
348
349 with the new behavior:
350
351 $ bison foo.y
352 foo.y: warning: 1 shift/reduce conflict [-Wconflicts-sr]
353 foo.y: warning: 2 reduce/reduce conflicts [-Wconflicts-rr]
354 $ bison -Werror foo.y
9503b0a4
TR
355 foo.y: error: 1 shift/reduce conflict [-Werror=conflicts-sr]
356 foo.y: error: 2 reduce/reduce conflicts [-Werror=conflicts-rr]
d87ea54c
AD
357
358 When %expect or %expect-rr is used, such as with bar.y:
359
360 %expect 0
361 %glr-parser
362 %%
363 exp: exp '+' exp | '0' | '0';
364
365 Former behavior:
366
367 $ bison bar.y
368 bar.y: conflicts: 1 shift/reduce, 2 reduce/reduce
369 bar.y: expected 0 shift/reduce conflicts
370 bar.y: expected 0 reduce/reduce conflicts
371
372 New one:
373
374 $ bison bar.y
9503b0a4
TR
375 bar.y: error: shift/reduce conflicts: 1 found, 0 expected
376 bar.y: error: reduce/reduce conflicts: 2 found, 0 expected
d87ea54c 377
f24695ef
AD
378** Incompatibilities with POSIX Yacc
379
380 The 'yacc' category is no longer part of '-Wall', enable it explicitly
381 with '-Wyacc'.
382
2055a44e
AD
383** Additional yylex/yyparse arguments
384
6dc4663d
AD
385 The new directive %param declares additional arguments to both yylex and
386 yyparse. The %lex-param, %parse-param, and %param directives support one
387 or more arguments. Instead of
2055a44e 388
e436fa67
AD
389 %lex-param {arg1_type *arg1}
390 %lex-param {arg2_type *arg2}
391 %parse-param {arg1_type *arg1}
392 %parse-param {arg2_type *arg2}
2055a44e
AD
393
394 one may now declare
395
e436fa67 396 %param {arg1_type *arg1} {arg2_type *arg2}
2055a44e 397
630a0218
AD
398** Types of values for %define variables
399
400 Bison used to make no difference between '%define foo bar' and '%define
401 foo "bar"'. The former is now called a 'keyword value', and the latter a
402 'string value'. A third kind was added: 'code values', such as '%define
403 foo {bar}'.
404
405 Keyword variables are used for fixed value sets, e.g.,
406
407 %define lr.type lalr
408
409 Code variables are used for value in the target language, e.g.,
410
411 %define api.value.type {struct semantic_type}
412
413 String variables are used remaining cases, e.g. file names.
414
2a6b66c5 415** Variable api.token.prefix
99c08fb6 416
2a6b66c5 417 The variable api.token.prefix changes the way tokens are identified in
99c08fb6
AD
418 the generated files. This is especially useful to avoid collisions
419 with identifiers in the target language. For instance
420
e436fa67 421 %token FILE for ERROR
630a0218 422 %define api.token.prefix {TOK_}
e436fa67
AD
423 %%
424 start: FILE for ERROR;
99c08fb6
AD
425
426 will generate the definition of the symbols TOK_FILE, TOK_for, and
427 TOK_ERROR in the generated sources. In particular, the scanner must
428 use these prefixed token names, although the grammar itself still
429 uses the short names (as in the sample rule given above).
430
6574576c
AD
431** Variable api.value.type
432
433 This new %define variable supersedes the #define macro YYSTYPE. The use
434 of YYSTYPE is discouraged. In particular, #defining YYSTYPE *and* either
435 using %union or %defining api.value.type results in undefined behavior.
436
437 Either define api.value.type, or use "%union":
438
439 %union
440 {
441 int ival;
442 char *sval;
443 }
444 %token <ival> INT "integer"
445 %token <sval> STRING "string"
446 %printer { fprintf (yyo, "%d", $$); } <ival>
447 %destructor { free ($$); } <sval>
448
449 /* In yylex(). */
450 yylval.ival = 42; return INT;
451 yylval.sval = "42"; return STRING;
452
630a0218
AD
453 The %define variable api.value.type supports both keyword and code values.
454
455 The keyword value 'union' means that the user provides genuine types, not
435575cb 456 union member names such as "ival" and "sval" above (WARNING: will fail if
1fa19a76 457 -y/--yacc/%yacc is enabled).
6574576c 458
435575cb 459 %define api.value.type union
6574576c
AD
460 %token <int> INT "integer"
461 %token <char *> STRING "string"
462 %printer { fprintf (yyo, "%d", $$); } <int>
463 %destructor { free ($$); } <char *>
464
465 /* In yylex(). */
466 yylval.INT = 42; return INT;
467 yylval.STRING = "42"; return STRING;
468
435575cb
AD
469 The keyword value variant is somewhat equivalent, but for C++ special
470 provision is made to allow classes to be used (more about this below).
6574576c 471
435575cb 472 %define api.value.type variant
6574576c
AD
473 %token <int> INT "integer"
474 %token <std::string> STRING "string"
475
630a0218 476 Code values (in braces) denote user defined types. This is where YYSTYPE
435575cb 477 used to be used.
6574576c
AD
478
479 %code requires
480 {
481 struct my_value
482 {
483 enum
484 {
485 is_int, is_string
486 } kind;
487 union
488 {
489 int ival;
490 char *sval;
491 } u;
492 };
493 }
435575cb 494 %define api.value.type {struct my_value}
6574576c
AD
495 %token <u.ival> INT "integer"
496 %token <u.sval> STRING "string"
497 %printer { fprintf (yyo, "%d", $$); } <u.ival>
498 %destructor { free ($$); } <u.sval>
499
500 /* In yylex(). */
501 yylval.u.ival = 42; return INT;
502 yylval.u.sval = "42"; return STRING;
503
31b850d2
AD
504** Variable parse.error
505
1f77b2e0
AD
506 This variable controls the verbosity of error messages. The use of the
507 %error-verbose directive is deprecated in favor of "%define parse.error
508 verbose".
31b850d2 509
c21e515e
AD
510** Renamed %define variables
511
512 The following variables have been renamed for consistency. Backward
513 compatibility is ensured, but upgrading is recommended.
514
515 lr.default-reductions -> lr.default-reduction
516 lr.keep-unreachable-states -> lr.keep-unreachable-state
517 namespace -> api.namespace
518 stype -> api.value.type
519
ca2a6d15
PH
520** Semantic predicates
521
597afd73
AD
522 Contributed by Paul Hilfinger.
523
1f77b2e0
AD
524 The new, experimental, semantic-predicate feature allows actions of the
525 form "%?{ BOOLEAN-EXPRESSION }", which cause syntax errors (as for
ca2a6d15 526 YYERROR) if the expression evaluates to 0, and are evaluated immediately
1f77b2e0
AD
527 in GLR parsers, rather than being deferred. The result is that they allow
528 the programmer to prune possible parses based on the values of run-time
529 expressions.
ca2a6d15 530
d1400569
AD
531** The directive %expect-rr is now an error in non GLR mode
532
533 It used to be an error only if used in non GLR mode, _and_ if there are
534 reduce/reduce conflicts.
535
5202b6ac 536** Tokens are numbered in their order of appearance
93561c21 537
5202b6ac
VT
538 Contributed by Valentin Tolmer.
539
540 With '%token A B', A had a number less than the one of B. However,
541 precedence declarations used to generate a reversed order. This is now
542 fixed, and introducing tokens with any of %token, %left, %right,
543 %precedence, or %nonassoc yields the same result.
544
545 When mixing declarations of tokens with a litteral character (e.g., 'a')
546 or with an identifier (e.g., B) in a precedence declaration, Bison
547 numbered the litteral characters first. For example
548
549 %right A B 'c' 'd'
550
551 would lead to the tokens declared in this order: 'c' 'd' A B. Again, the
552 input order is now preserved.
553
554 These changes were made so that one can remove useless precedence and
555 associativity declarations (i.e., map %nonassoc, %left or %right to
556 %precedence, or to %token) and get exactly the same output.
93561c21 557
cc2235ac
VT
558** Useless precedence and associativity
559
d2f9ae18
AD
560 Contributed by Valentin Tolmer.
561
1282c124 562 When developing and maintaining a grammar, useless associativity and
cc2235ac
VT
563 precedence directives are common. They can be a nuisance: new ambiguities
564 arising are sometimes masked because their conflicts are resolved due to
565 the extra precedence or associativity information. Furthermore, it can
566 hinder the comprehension of a new grammar: one will wonder about the role
567 of a precedence, where in fact it is useless. The following changes aim
568 at detecting and reporting these extra directives.
569
570*** Precedence warning category
571
572 A new category of warning, -Wprecedence, was introduced. It flags the
573 useless precedence and associativity directives.
574
575*** Useless associativity
576
577 Bison now warns about symbols with a declared associativity that is never
578 used to resolve conflicts. In that case, using %precedence is sufficient;
579 the parsing tables will remain unchanged. Solving these warnings may raise
580 useless precedence warnings, as the symbols no longer have associativity.
581 For example:
582
583 %left '+'
584 %left '*'
585 %%
586 exp:
1282c124
AD
587 "number"
588 | exp '+' "number"
cc2235ac
VT
589 | exp '*' exp
590 ;
591
592 will produce a
593
594 warning: useless associativity for '+', use %precedence [-Wprecedence]
595 %left '+'
596 ^^^
597
598*** Useless precedence
599
600 Bison now warns about symbols with a declared precedence and no declared
601 associativity (i.e., declared with %precedence), and whose precedence is
602 never used. In that case, the symbol can be safely declared with %token
603 instead, without modifying the parsing tables. For example:
604
605 %precedence '='
606 %%
1282c124 607 exp: "var" '=' "number";
cc2235ac
VT
608
609 will produce a
610
611 warning: useless precedence for '=' [-Wprecedence]
612 %precedence '='
613 ^^^
614
615*** Useless precedence and associativity
616
617 In case of both useless precedence and associativity, the issue is flagged
618 as follows:
619
620 %nonassoc '='
621 %%
1282c124 622 exp: "var" '=' "number";
cc2235ac
VT
623
624 The warning is:
625
626 warning: useless precedence and associativity for '=' [-Wprecedence]
627 %nonassoc '='
628 ^^^
09add9c2
AD
629
630** Empty rules
631
6240346a
AD
632 With help from Joel E. Denny and Gabriel Rassoul.
633
09add9c2
AD
634 Empty rules (i.e., with an empty right-hand side) can now be explicitly
635 marked by the new %empty directive. Using %empty on a non-empty rule is
636 an error. The new -Wempty-rule warning reports empty rules without
637 %empty. On the following grammar:
638
639 %%
640 s: a b c;
641 a: ;
642 b: %empty;
643 c: 'a' %empty;
644
645 bison reports:
646
647 3.4-5: warning: empty rule without %empty [-Wempty-rule]
648 a: {}
649 ^^
650 5.8-13: error: %empty on non-empty rule
651 c: 'a' %empty {};
652 ^^^^^^
cc2235ac 653
c21e515e
AD
654** Java skeleton improvements
655
c21e515e
AD
656 The constants for token names were moved to the Lexer interface. Also, it
657 is possible to add code to the parser's constructors using "%code init"
658 and "%define init_throws".
aa94def1
DH
659 Contributed by Paolo Bonzini.
660
de1a2f20
AD
661 The Java skeleton now supports push parsing.
662 Contributed by Dennis Heimbigner.
663
c21e515e
AD
664** C++ skeletons improvements
665
666*** The parser header is no longer mandatory (lalr1.cc, glr.cc)
667
668 Using %defines is now optional. Without it, the needed support classes
669 are defined in the generated parser, instead of additional files (such as
670 location.hh, position.hh and stack.hh).
671
672*** Locations are no longer mandatory (lalr1.cc, glr.cc)
673
674 Both lalr1.cc and glr.cc no longer require %location.
675
676*** syntax_error exception (lalr1.cc)
677
678 The C++ parser features a syntax_error exception, which can be
679 thrown from the scanner or from user rules to raise syntax errors.
680 This facilitates reporting errors caught in sub-functions (e.g.,
681 rejecting too large integral literals from a conversion function
682 used by the scanner, or rejecting invalid combinations from a
683 factory invoked by the user actions).
684
685*** %define api.value.type variant
686
687 This is based on a submission from Michiel De Wilde. With help
688 from Théophile Ranquet.
689
690 In this mode, complex C++ objects can be used as semantic values. For
691 instance:
692
693 %token <::std::string> TEXT;
694 %token <int> NUMBER;
695 %token SEMICOLON ";"
696 %type <::std::string> item;
697 %type <::std::list<std::string>> list;
698 %%
699 result:
700 list { std::cout << $1 << std::endl; }
701 ;
702
703 list:
6240346a 704 %empty { /* Generates an empty string list. */ }
c21e515e
AD
705 | list item ";" { std::swap ($$, $1); $$.push_back ($2); }
706 ;
707
708 item:
709 TEXT { std::swap ($$, $1); }
710 | NUMBER { $$ = string_cast ($1); }
711 ;
712
713*** %define api.token.constructor
714
715 When variants are enabled, Bison can generate functions to build the
716 tokens. This guarantees that the token type (e.g., NUMBER) is consistent
717 with the semantic value (e.g., int):
718
719 parser::symbol_type yylex ()
720 {
721 parser::location_type loc = ...;
722 ...
723 return parser::make_TEXT ("Hello, world!", loc);
724 ...
725 return parser::make_NUMBER (42, loc);
726 ...
727 return parser::make_SEMICOLON (loc);
728 ...
729 }
730
75ae8299
AD
731*** C++ locations
732
733 There are operator- and operator-= for 'location'. Negative line/column
734 increments can no longer underflow the resulting value.
735
1f5542fe 736* Noteworthy changes in release 2.7.1 (2013-04-15) [stable]
cc8962bd 737
80a2826e
AD
738** Bug fixes
739
740*** Fix compiler attribute portability (yacc.c)
741
742 With locations enabled, __attribute__ was used unprotected.
0a7b8559 743
e83be476
AD
744*** Fix some compiler warnings (lalr1.cc)
745
c13bb348 746* Noteworthy changes in release 2.7 (2012-12-12) [stable]
effd30c0 747
edf9a06f 748** Bug fixes
7bada535 749
edf9a06f 750 Warnings about uninitialized yylloc in yyparse have been fixed.
7bada535 751
1127a75a
AD
752 Restored C90 compliance (yet no report was ever made).
753
d4fe9e88 754** Diagnostics are improved
7bada535 755
8458a411
AD
756 Contributed by Théophile Ranquet.
757
d4fe9e88 758*** Changes in the format of error messages
7bada535 759
d4fe9e88 760 This used to be the format of many error reports:
1f1bd572 761
d4fe9e88
AD
762 input.y:2.7-12: %type redeclaration for exp
763 input.y:1.7-12: previous declaration
1f1bd572 764
d4fe9e88 765 It is now:
1f1bd572 766
d4fe9e88
AD
767 input.y:2.7-12: error: %type redeclaration for exp
768 input.y:1.7-12: previous declaration
cbaea010 769
d4fe9e88 770*** New format for error reports: carets
cbaea010 771
d4fe9e88 772 Caret errors have been added to Bison:
cbaea010 773
d4fe9e88
AD
774 input.y:2.7-12: error: %type redeclaration for exp
775 %type <sval> exp
776 ^^^^^^
777 input.y:1.7-12: previous declaration
778 %type <ival> exp
779 ^^^^^^
cbaea010 780
d4fe9e88 781 or
cbaea010 782
7bada535 783 input.y:3.20-23: error: ambiguous reference: '$exp'
fb6040f0 784 exp: exp '+' exp { $exp = $1 + $3; };
7bada535 785 ^^^^
fb6040f0
TR
786 input.y:3.1-3: refers to: $exp at $$
787 exp: exp '+' exp { $exp = $1 + $3; };
788 ^^^
789 input.y:3.6-8: refers to: $exp at $1
790 exp: exp '+' exp { $exp = $1 + $3; };
791 ^^^
792 input.y:3.14-16: refers to: $exp at $3
793 exp: exp '+' exp { $exp = $1 + $3; };
794 ^^^
7bada535 795
1282c124
AD
796 The default behavior for now is still not to display these unless
797 explicitly asked with -fcaret (or -fall). However, in a later release, it
d4fe9e88
AD
798 will be made the default behavior (but may still be deactivated with
799 -fno-caret).
d3e4409a 800
1f1bd572 801** New value for %define variable: api.pure full
d3e4409a 802
1f1bd572 803 The %define variable api.pure requests a pure (reentrant) parser. However,
d4fe9e88
AD
804 for historical reasons, using it in a location-tracking Yacc parser
805 resulted in a yyerror function that did not take a location as a
806 parameter. With this new value, the user may request a better pure parser,
807 where yyerror does take a location as a parameter (in location-tracking
808 parsers).
1f1bd572
TR
809
810 The use of "%define api.pure true" is deprecated in favor of this new
811 "%define api.pure full".
d3e4409a 812
7287be84 813** New %define variable: api.location.type (glr.cc, lalr1.cc, lalr1.java)
db8ab2be
AD
814
815 The %define variable api.location.type defines the name of the type to use
816 for locations. When defined, Bison no longer generates the position.hh
817 and location.hh files, nor does the parser will include them: the user is
818 then responsible to define her type.
819
820 This can be used in programs with several parsers to factor their location
7287be84
AD
821 and position files: let one of them generate them, and the others just use
822 them.
db8ab2be
AD
823
824 This feature was actually introduced, but not documented, in Bison 2.5,
825 under the name "location_type" (which is maintained for backward
826 compatibility).
827
7287be84
AD
828 For consistency, lalr1.java's %define variables location_type and
829 position_type are deprecated in favor of api.location.type and
830 api.position.type.
831
d4fe9e88
AD
832** Exception safety (lalr1.cc)
833
834 The parse function now catches exceptions, uses the %destructors to
835 release memory (the lookahead symbol and the symbols pushed on the stack)
836 before re-throwing the exception.
837
838 This feature is somewhat experimental. User feedback would be
839 appreciated.
840
9c16d399 841** Graph improvements in DOT and XSLT
fc4fdd62 842
8458a411
AD
843 Contributed by Théophile Ranquet.
844
fc4fdd62
TR
845 The graphical presentation of the states is more readable: their shape is
846 now rectangular, the state number is clearly displayed, and the items are
847 numbered and left-justified.
848
849 The reductions are now explicitly represented as transitions to other
850 diamond shaped nodes.
851
9c16d399
TR
852 These changes are present in both --graph output and xml2dot.xsl XSLT
853 processing, with minor (documented) differences.
854
d4fe9e88 855** %language is no longer an experimental feature.
fb4c8a7c 856
d4fe9e88
AD
857 The introduction of this feature, in 2.4, was four years ago. The
858 --language option and the %language directive are no longer experimental.
fb4c8a7c 859
53e2cd1e
AD
860** Documentation
861
862 The sections about shift/reduce and reduce/reduce conflicts resolution
863 have been fixed and extended.
9d2423f5 864
d4fe9e88
AD
865 Although introduced more than four years ago, XML and Graphviz reports
866 were not properly documented.
867
be22823e
AD
868 The translation of mid-rule actions is now described.
869
9d3f7eaf 870* Noteworthy changes in release 2.6.5 (2012-11-07) [stable]
6f1360bd 871
a68b1f23
AD
872 We consider compiler warnings about Bison generated parsers to be bugs.
873 Rather than working around them in your own project, please consider
874 reporting them to us.
875
876** Bug fixes
877
878 Warnings about uninitialized yylval and/or yylloc for push parsers with a
879 pure interface have been fixed for GCC 4.0 up to 4.8, and Clang 2.9 to
880 3.2.
881
882 Other issues in the test suite have been addressed.
6f1360bd 883
1282c124 884 Null characters are correctly displayed in error messages.
95066e92 885
a1d1ab50
AD
886 When possible, yylloc is correctly initialized before calling yylex. It
887 is no longer necessary to initialize it in the %initial-action.
888
0ac15849 889* Noteworthy changes in release 2.6.4 (2012-10-23) [stable]
a4eb820f 890
468455e1 891 Bison 2.6.3's --version was incorrect. This release fixes this issue.
a4eb820f 892
6eb8f74f 893* Noteworthy changes in release 2.6.3 (2012-10-22) [stable]
933ec544 894
6b4cb804
AD
895** Bug fixes
896
a1a77e1f 897 Bugs and portability issues in the test suite have been fixed.
6b4cb804
AD
898
899 Some errors in translations have been addressed, and --help now directs
900 users to the appropriate place to report them.
901
902 Stray Info files shipped by accident are removed.
903
904 Incorrect definitions of YY_, issued by yacc.c when no parser header is
905 generated, are removed.
9c26b8fc 906
a2b3f101
TR
907 All the generated headers are self-contained.
908
c9d5bcc9
AD
909** Header guards (yacc.c, glr.c, glr.cc)
910
911 In order to avoid collisions, the header guards are now
912 YY_<PREFIX>_<FILE>_INCLUDED, instead of merely <PREFIX>_<FILE>.
913 For instance the header generated from
914
915 %define api.prefix "calc"
916 %defines "lib/parse.h"
917
918 will use YY_CALC_LIB_PARSE_H_INCLUDED as guard.
919
c12c4c50 920** Fix compiler warnings in the generated parser (yacc.c, glr.c)
321d3e35
AD
921
922 The compilation of pure parsers (%define api.pure) can trigger GCC
923 warnings such as:
924
925 input.c: In function 'yyparse':
926 input.c:1503:12: warning: 'yylval' may be used uninitialized in this
927 function [-Wmaybe-uninitialized]
928 *++yyvsp = yylval;
929 ^
930
931 This is now fixed; pragmas to avoid these warnings are no longer needed.
932
c12c4c50
AD
933 Warnings from clang ("equality comparison with extraneous parentheses" and
934 "function declared 'noreturn' should not return") have also been
935 addressed.
936
e1eeecd3 937* Noteworthy changes in release 2.6.2 (2012-08-03) [stable]
9c26b8fc 938
43ca8040
AD
939** Bug fixes
940
941 Buffer overruns, complaints from Flex, and portability issues in the test
942 suite have been fixed.
943
c9d546b2
AD
944** Spaces in %lex- and %parse-param (lalr1.cc, glr.cc)
945
946 Trailing end-of-lines in %parse-param or %lex-param would result in
947 invalid C++. This is fixed.
9c26b8fc 948
dcd5344d
AD
949** Spurious spaces and end-of-lines
950
951 The generated files no longer end (nor start) with empty lines.
952
77b214ef 953* Noteworthy changes in release 2.6.1 (2012-07-30) [stable]
a4107f24 954
8617d87e
AD
955 Bison no longer executes user-specified M4 code when processing a grammar.
956
e20e6a50
AD
957** Future Changes
958
959 In addition to the removal of the features announced in Bison 2.6, the
960 next major release will remove the "Temporary hack for adding a semicolon
961 to the user action", as announced in the release 2.5. Instead of:
962
963 exp: exp "+" exp { $$ = $1 + $3 };
964
965 write:
966
967 exp: exp "+" exp { $$ = $1 + $3; };
968
8617d87e
AD
969** Bug fixes
970
0e164d43
AD
971*** Type names are now properly escaped.
972
973*** glr.cc: set_debug_level and debug_level work as expected.
a4107f24 974
26313726
AD
975*** Stray @ or $ in actions
976
977 While Bison used to warn about stray $ or @ in action rules, it did not
978 for other actions such as printers, destructors, or initial actions. It
979 now does.
980
cd735a8c 981** Type names in actions
4982f078
AD
982
983 For consistency with rule actions, it is now possible to qualify $$ by a
cd735a8c 984 type-name in destructors, printers, and initial actions. For instance:
4982f078
AD
985
986 %printer { fprintf (yyo, "(%d, %f)", $<ival>$, $<fval>$); } <*> <>;
987
988 will display two values for each typed and untyped symbol (provided
cd735a8c 989 that YYSTYPE has both "ival" and "fval" fields).
60aa04a2 990
1505e8bb 991* Noteworthy changes in release 2.6 (2012-07-19) [stable]
0f11eec2 992
d0a30438 993** Future changes
9553083c 994
55d1006f
AD
995 The next major release of Bison will drop support for the following
996 deprecated features. Please report disagreements to bug-bison@gnu.org.
0f11eec2 997
aaf61036 998*** K&R C parsers
55d1006f
AD
999
1000 Support for generating parsers in K&R C will be removed. Parsers
242cc08e 1001 generated for C support ISO C90, and are tested with ISO C99 and ISO C11
55d1006f
AD
1002 compilers.
1003
258cddbc 1004*** Features deprecated since Bison 1.875
0f11eec2 1005
258cddbc
AD
1006 The definitions of yystype and yyltype will be removed; use YYSTYPE and
1007 YYLTYPE.
0f11eec2 1008
258cddbc
AD
1009 YYPARSE_PARAM and YYLEX_PARAM, deprecated in favor of %parse-param and
1010 %lex-param, will no longer be supported.
1011
1012 Support for the preprocessor symbol YYERROR_VERBOSE will be removed, use
1013 %error-verbose.
55d1006f
AD
1014
1015*** The generated header will be included (yacc.c)
0f11eec2
AD
1016
1017 Instead of duplicating the content of the generated header (definition of
55d1006f
AD
1018 YYSTYPE, yyparse declaration etc.), the generated parser will include it,
1019 as is already the case for GLR or C++ parsers. This change is deferred
1020 because existing versions of ylwrap (e.g., Automake 1.12.1) do not support
1021 it.
0f11eec2 1022
c2425191 1023** Generated Parser Headers
56ca3d8f 1024
258cddbc 1025*** Guards (yacc.c, glr.c, glr.cc)
c3e9f08f
AD
1026
1027 The generated headers are now guarded, as is already the case for C++
242cc08e 1028 parsers (lalr1.cc). For instance, with --defines=foo.h:
c3e9f08f 1029
e29f0771
AD
1030 #ifndef YY_FOO_H
1031 # define YY_FOO_H
1032 ...
1033 #endif /* !YY_FOO_H */
c3e9f08f 1034
258cddbc 1035*** New declarations (yacc.c, glr.c)
56ca3d8f
AD
1036
1037 The generated header now declares yydebug and yyparse. Both honor
1038 --name-prefix=bar_, and yield
1039
e29f0771 1040 int bar_parse (void);
56ca3d8f
AD
1041
1042 rather than
1043
e29f0771
AD
1044 #define yyparse bar_parse
1045 int yyparse (void);
56ca3d8f
AD
1046
1047 in order to facilitate the inclusion of several parser headers inside a
1048 single compilation unit.
c3e9f08f 1049
258cddbc
AD
1050*** Exported symbols in C++
1051
1052 The symbols YYTOKEN_TABLE and YYERROR_VERBOSE, which were defined in the
1053 header, are removed, as they prevent the possibility of including several
1054 generated headers from a single compilation unit.
1055
694af10c
AD
1056*** YYLSP_NEEDED
1057
1058 For the same reasons, the undocumented and unused macro YYLSP_NEEDED is no
1059 longer defined.
1060
4b3847c3
AD
1061** New %define variable: api.prefix
1062
1063 Now that the generated headers are more complete and properly protected
1064 against multiple inclusions, constant names, such as YYSTYPE are a
1065 problem. While yyparse and others are properly renamed by %name-prefix,
1066 YYSTYPE, YYDEBUG and others have never been affected by it. Because it
1067 would introduce backward compatibility issues in projects not expecting
1068 YYSTYPE to be renamed, instead of changing the behavior of %name-prefix,
1069 it is deprecated in favor of a new %define variable: api.prefix.
1070
1071 The following examples compares both:
1072
1073 %name-prefix "bar_" | %define api.prefix "bar_"
1074 %token <ival> FOO %token <ival> FOO
1075 %union { int ival; } %union { int ival; }
1076 %% %%
1077 exp: 'a'; exp: 'a';
1078
1079 bison generates:
1080
1081 #ifndef BAR_FOO_H #ifndef BAR_FOO_H
1082 # define BAR_FOO_H # define BAR_FOO_H
1083
1084 /* Enabling traces. */ /* Enabling traces. */
5f108727
AD
1085 # ifndef YYDEBUG | # ifndef BAR_DEBUG
1086 > # if defined YYDEBUG
1087 > # if YYDEBUG
1088 > # define BAR_DEBUG 1
1089 > # else
1090 > # define BAR_DEBUG 0
1091 > # endif
1092 > # else
1093 # define YYDEBUG 0 | # define BAR_DEBUG 0
1094 > # endif
1095 # endif | # endif
1096
1097 # if YYDEBUG | # if BAR_DEBUG
4b3847c3
AD
1098 extern int bar_debug; extern int bar_debug;
1099 # endif # endif
1100
1101 /* Tokens. */ /* Tokens. */
1102 # ifndef YYTOKENTYPE | # ifndef BAR_TOKENTYPE
1103 # define YYTOKENTYPE | # define BAR_TOKENTYPE
1104 enum yytokentype { | enum bar_tokentype {
1105 FOO = 258 FOO = 258
1106 }; };
1107 # endif # endif
1108
1109 #if ! defined YYSTYPE \ | #if ! defined BAR_STYPE \
1110 && ! defined YYSTYPE_IS_DECLARED | && ! defined BAR_STYPE_IS_DECLARED
1111 typedef union YYSTYPE | typedef union BAR_STYPE
1112 { {
1113 int ival; int ival;
1114 } YYSTYPE; | } BAR_STYPE;
1115 # define YYSTYPE_IS_DECLARED 1 | # define BAR_STYPE_IS_DECLARED 1
1116 #endif #endif
1117
1118 extern YYSTYPE bar_lval; | extern BAR_STYPE bar_lval;
1119
1120 int bar_parse (void); int bar_parse (void);
1121
1122 #endif /* !BAR_FOO_H */ #endif /* !BAR_FOO_H */
1123
dfaac272 1124* Noteworthy changes in release 2.5.1 (2012-06-05) [stable]
df6e3db0 1125
debe2c03 1126** Future changes:
765e1bd4 1127
e4ab1254 1128 The next major release will drop support for generating parsers in K&R C.
041308d0 1129
466b4cf2 1130** yacc.c: YYBACKUP works as expected.
ef51bfa7 1131
d834eca0 1132** glr.c improvements:
041308d0 1133
d834eca0 1134*** Location support is eliminated when not requested:
041308d0 1135
e4ab1254
AD
1136 GLR parsers used to include location-related code even when locations were
1137 not requested, and therefore not even usable.
378e917c 1138
d834eca0 1139*** __attribute__ is preserved:
d115aad9 1140
e4ab1254
AD
1141 __attribute__ is no longer disabled when __STRICT_ANSI__ is defined (i.e.,
1142 when -std is passed to GCC).
041308d0 1143
466b4cf2 1144** lalr1.java: several fixes:
041308d0 1145
e4ab1254
AD
1146 The Java parser no longer throws ArrayIndexOutOfBoundsException if the
1147 first token leads to a syntax error. Some minor clean ups.
041308d0 1148
22172d47 1149** Changes for C++:
ef51bfa7 1150
22172d47 1151*** C++11 compatibility:
ef51bfa7 1152
e4ab1254
AD
1153 C and C++ parsers use "nullptr" instead of "0" when __cplusplus is 201103L
1154 or higher.
936c88d1 1155
22172d47
AD
1156*** Header guards
1157
1158 The header files such as "parser.hh", "location.hh", etc. used a constant
1159 name for preprocessor guards, for instance:
1160
e29f0771
AD
1161 #ifndef BISON_LOCATION_HH
1162 # define BISON_LOCATION_HH
1163 ...
1164 #endif // !BISON_LOCATION_HH
22172d47
AD
1165
1166 The inclusion guard is now computed from "PREFIX/FILE-NAME", where lower
1167 case characters are converted to upper case, and series of
1168 non-alphanumerical characters are converted to an underscore.
1169
1170 With "bison -o lang++/parser.cc", "location.hh" would now include:
1171
e29f0771
AD
1172 #ifndef YY_LANG_LOCATION_HH
1173 # define YY_LANG_LOCATION_HH
1174 ...
1175 #endif // !YY_LANG_LOCATION_HH
22172d47
AD
1176
1177*** C++ locations:
936c88d1 1178
e4ab1254
AD
1179 The position and location constructors (and their initialize methods)
1180 accept new arguments for line and column. Several issues in the
1181 documentation were fixed.
936c88d1 1182
466b4cf2
AD
1183** liby is no longer asking for "rpl_fprintf" on some platforms.
1184
7e508a2b
AD
1185** Changes in the manual:
1186
1187*** %printer is documented
1188
e4ab1254
AD
1189 The "%printer" directive, supported since at least Bison 1.50, is finally
1190 documented. The "mfcalc" example is extended to demonstrate it.
7e508a2b 1191
e4ab1254
AD
1192 For consistency with the C skeletons, the C++ parsers now also support
1193 "yyoutput" (as an alias to "debug_stream ()").
7e508a2b
AD
1194
1195*** Several improvements have been made:
466b4cf2 1196
e4ab1254
AD
1197 The layout for grammar excerpts was changed to a more compact scheme.
1198 Named references are motivated. The description of the automaton
1199 description file (*.output) is updated to the current format. Incorrect
1200 index entries were fixed. Some other errors were fixed.
466b4cf2 1201
86b08b49
AD
1202** Building bison:
1203
1204*** Conflicting prototypes with recent/modified Flex.
466b4cf2 1205
e4ab1254
AD
1206 Fixed build problems with the current, unreleased, version of Flex, and
1207 some modified versions of 2.5.35, which have modified function prototypes.
466b4cf2 1208
8ef26c2a
AD
1209*** Warnings during the build procedure have been eliminated.
1210
1211*** Several portability problems in the test suite have been fixed:
466b4cf2 1212
e4ab1254
AD
1213 This includes warnings with some compilers, unexpected behavior of tools
1214 such as diff, warning messages from the test suite itself, etc.
466b4cf2 1215
91aadcc7 1216*** The install-pdf target works properly:
8ef26c2a 1217
e4ab1254
AD
1218 Running "make install-pdf" (or -dvi, -html, -info, and -ps) no longer
1219 halts in the middle of its course.
8ef26c2a 1220
28801043 1221* Changes in version 2.5 (2011-05-14):
50cca368 1222
82f3355e
JD
1223** Grammar symbol names can now contain non-initial dashes:
1224
1225 Consistently with directives (such as %error-verbose) and with
1226 %define variables (e.g. push-pull), grammar symbol names may contain
1227 dashes in any position except the beginning. This is a GNU
1228 extension over POSIX Yacc. Thus, use of this extension is reported
1229 by -Wyacc and rejected in Yacc mode (--yacc).
1230
f1b238df 1231** Named references:
66381412
AR
1232
1233 Historically, Yacc and Bison have supported positional references
1234 ($n, $$) to allow access to symbol values from inside of semantic
1235 actions code.
1236
1237 Starting from this version, Bison can also accept named references.
1238 When no ambiguity is possible, original symbol names may be used
1239 as named references:
1240
4b568fc0 1241 if_stmt : "if" cond_expr "then" then_stmt ';'
66381412
AR
1242 { $if_stmt = mk_if_stmt($cond_expr, $then_stmt); }
1243
1244 In the more common case, explicit names may be declared:
1245
4b568fc0 1246 stmt[res] : "if" expr[cond] "then" stmt[then] "else" stmt[else] ';'
66381412
AR
1247 { $res = mk_if_stmt($cond, $then, $else); }
1248
5b1ff423 1249 Location information is also accessible using @name syntax. When
66381412
AR
1250 accessing symbol names containing dots or dashes, explicit bracketing
1251 ($[sym.1]) must be used.
1252
5b1ff423 1253 These features are experimental in this version. More user feedback
66381412 1254 will help to stabilize them.
2bd435c3 1255 Contributed by Alex Rozenman.
66381412 1256
f1b238df 1257** IELR(1) and canonical LR(1):
eb45ef3b
JD
1258
1259 IELR(1) is a minimal LR(1) parser table generation algorithm. That
1260 is, given any context-free grammar, IELR(1) generates parser tables
7262f54f 1261 with the full language-recognition power of canonical LR(1) but with
f1b238df
JD
1262 nearly the same number of parser states as LALR(1). This reduction
1263 in parser states is often an order of magnitude. More importantly,
eb45ef3b
JD
1264 because canonical LR(1)'s extra parser states may contain duplicate
1265 conflicts in the case of non-LR(1) grammars, the number of conflicts
1266 for IELR(1) is often an order of magnitude less as well. This can
1267 significantly reduce the complexity of developing of a grammar.
1268
1269 Bison can now generate IELR(1) and canonical LR(1) parser tables in
1270 place of its traditional LALR(1) parser tables, which remain the
1271 default. You can specify the type of parser tables in the grammar
1272 file with these directives:
1273
cf499cff
JD
1274 %define lr.type lalr
1275 %define lr.type ielr
1276 %define lr.type canonical-lr
eb45ef3b 1277
7fceb615 1278 The default-reduction optimization in the parser tables can also be
e4ab1254
AD
1279 adjusted using "%define lr.default-reductions". For details on both
1280 of these features, see the new section "Tuning LR" in the Bison
7fceb615 1281 manual.
eb45ef3b
JD
1282
1283 These features are experimental. More user feedback will help to
1284 stabilize them.
1285
8458a411
AD
1286** LAC (Lookahead Correction) for syntax error handling
1287
1288 Contributed by Joel E. Denny.
fcf834f9
JD
1289
1290 Canonical LR, IELR, and LALR can suffer from a couple of problems
1291 upon encountering a syntax error. First, the parser might perform
1292 additional parser stack reductions before discovering the syntax
7fceb615 1293 error. Such reductions can perform user semantic actions that are
fcf834f9
JD
1294 unexpected because they are based on an invalid token, and they
1295 cause error recovery to begin in a different syntactic context than
1296 the one in which the invalid token was encountered. Second, when
7fceb615 1297 verbose error messages are enabled (with %error-verbose or the
e4ab1254 1298 obsolete "#define YYERROR_VERBOSE"), the expected token list in the
7fceb615
JD
1299 syntax error message can both contain invalid tokens and omit valid
1300 tokens.
fcf834f9
JD
1301
1302 The culprits for the above problems are %nonassoc, default
1303 reductions in inconsistent states, and parser state merging. Thus,
1304 IELR and LALR suffer the most. Canonical LR can suffer only if
1305 %nonassoc is used or if default reductions are enabled for
1306 inconsistent states.
1307
7fceb615
JD
1308 LAC is a new mechanism within the parsing algorithm that solves
1309 these problems for canonical LR, IELR, and LALR without sacrificing
1310 %nonassoc, default reductions, or state merging. When LAC is in
1311 use, canonical LR and IELR behave almost exactly the same for both
1312 syntactically acceptable and syntactically unacceptable input.
fcf834f9
JD
1313 While LALR still does not support the full language-recognition
1314 power of canonical LR and IELR, LAC at least enables LALR's syntax
1315 error handling to correctly reflect LALR's language-recognition
1316 power.
1317
1318 Currently, LAC is only supported for deterministic parsers in C.
1319 You can enable LAC with the following directive:
1320
1321 %define parse.lac full
1322
e4ab1254 1323 See the new section "LAC" in the Bison manual for additional
7fceb615 1324 details including a few caveats.
fcf834f9
JD
1325
1326 LAC is an experimental feature. More user feedback will help to
1327 stabilize it.
1328
d397d9f0 1329** %define improvements:
cf499cff 1330
f1b238df 1331*** Can now be invoked via the command line:
50cca368 1332
de5ab940 1333 Each of these command-line options
50cca368 1334
de5ab940
JD
1335 -D NAME[=VALUE]
1336 --define=NAME[=VALUE]
1337
1338 -F NAME[=VALUE]
1339 --force-define=NAME[=VALUE]
50cca368
JD
1340
1341 is equivalent to this grammar file declaration
1342
de5ab940 1343 %define NAME ["VALUE"]
50cca368 1344
de5ab940
JD
1345 except that the manner in which Bison processes multiple definitions
1346 for the same NAME differs. Most importantly, -F and --force-define
1347 quietly override %define, but -D and --define do not. For further
e4ab1254 1348 details, see the section "Bison Options" in the Bison manual.
50cca368 1349
f1b238df 1350*** Variables renamed:
67212941
JD
1351
1352 The following %define variables
1353
1354 api.push_pull
1355 lr.keep_unreachable_states
1356
1357 have been renamed to
1358
1359 api.push-pull
1360 lr.keep-unreachable-states
1361
1362 The old names are now deprecated but will be maintained indefinitely
1363 for backward compatibility.
1364
7262f54f 1365*** Values no longer need to be quoted in the grammar file:
cf499cff
JD
1366
1367 If a %define value is an identifier, it no longer needs to be placed
1368 within quotations marks. For example,
1369
1370 %define api.push-pull "push"
1371
1372 can be rewritten as
1373
1374 %define api.push-pull push
1375
d397d9f0 1376*** Unrecognized variables are now errors not warnings.
cdf3f113 1377
d397d9f0
JD
1378*** Multiple invocations for any variable is now an error not a warning.
1379
1380** Unrecognized %code qualifiers are now errors not warnings.
1381
1382** Character literals not of length one:
1383
1384 Previously, Bison quietly converted all character literals to length
1385 one. For example, without warning, Bison interpreted the operators in
1386 the following grammar to be the same token:
1387
1388 exp: exp '++'
1389 | exp '+' exp
1390 ;
1391
1392 Bison now warns when a character literal is not of length one. In
1393 some future release, Bison will start reporting an error instead.
1394
1395** Destructor calls fixed for lookaheads altered in semantic actions:
1396
1397 Previously for deterministic parsers in C, if a user semantic action
1398 altered yychar, the parser in some cases used the old yychar value to
1399 determine which destructor to call for the lookahead upon a syntax
1400 error or upon parser return. This bug has been fixed.
1401
1402** C++ parsers use YYRHSLOC:
1403
1404 Similarly to the C parsers, the C++ parsers now define the YYRHSLOC
1405 macro and use it in the default YYLLOC_DEFAULT. You are encouraged
e4ab1254
AD
1406 to use it. If, for instance, your location structure has "first"
1407 and "last" members, instead of
d397d9f0 1408
e29f0771
AD
1409 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1410 do \
1411 if (N) \
1412 { \
1413 (Current).first = (Rhs)[1].location.first; \
1414 (Current).last = (Rhs)[N].location.last; \
1415 } \
1416 else \
1417 { \
1418 (Current).first = (Current).last = (Rhs)[0].location.last; \
1419 } \
1420 while (false)
d397d9f0
JD
1421
1422 use:
1423
e29f0771
AD
1424 # define YYLLOC_DEFAULT(Current, Rhs, N) \
1425 do \
1426 if (N) \
1427 { \
1428 (Current).first = YYRHSLOC (Rhs, 1).first; \
1429 (Current).last = YYRHSLOC (Rhs, N).last; \
1430 } \
1431 else \
1432 { \
1433 (Current).first = (Current).last = YYRHSLOC (Rhs, 0).last; \
1434 } \
1435 while (false)
d397d9f0
JD
1436
1437** YYLLOC_DEFAULT in C++:
1438
1439 The default implementation of YYLLOC_DEFAULT used to be issued in
1440 the header file. It is now output in the implementation file, after
1441 the user %code sections so that its #ifndef guard does not try to
1442 override the user's YYLLOC_DEFAULT if provided.
cdf3f113 1443
f1b238df 1444** YYFAIL now produces warnings and Java parsers no longer implement it:
4395a9ff
JD
1445
1446 YYFAIL has existed for many years as an undocumented feature of
1447 deterministic parsers in C generated by Bison. More recently, it was
1448 a documented feature of Bison's experimental Java parsers. As
1449 promised in Bison 2.4.2's NEWS entry, any appearance of YYFAIL in a
1450 semantic action now produces a deprecation warning, and Java parsers
1451 no longer implement YYFAIL at all. For further details, including a
1452 discussion of how to suppress C preprocessor warnings about YYFAIL
1453 being unused, see the Bison 2.4.2 NEWS entry.
1454
f1b238df 1455** Temporary hack for adding a semicolon to the user action:
197b82ba
JD
1456
1457 Previously, Bison appended a semicolon to every user action for
1458 reductions when the output language defaulted to C (specifically, when
1459 neither %yacc, %language, %skeleton, or equivalent command-line
1460 options were specified). This allowed actions such as
1461
1462 exp: exp "+" exp { $$ = $1 + $3 };
1463
1464 instead of
1465
1466 exp: exp "+" exp { $$ = $1 + $3; };
1467
1468 As a first step in removing this misfeature, Bison now issues a
1469 warning when it appends a semicolon. Moreover, in cases where Bison
1470 cannot easily determine whether a semicolon is needed (for example, an
1471 action ending with a cpp directive or a braced compound initializer),
1472 it no longer appends one. Thus, the C compiler might now complain
1473 about a missing semicolon where it did not before. Future releases of
1474 Bison will cease to append semicolons entirely.
1475
d2060f06
JD
1476** Verbose syntax error message fixes:
1477
e4ab1254 1478 When %error-verbose or the obsolete "#define YYERROR_VERBOSE" is
7fceb615
JD
1479 specified, syntax error messages produced by the generated parser
1480 include the unexpected token as well as a list of expected tokens.
1481 The effect of %nonassoc on these verbose messages has been corrected
1482 in two ways, but a more complete fix requires LAC, described above:
d2060f06
JD
1483
1484*** When %nonassoc is used, there can exist parser states that accept no
1485 tokens, and so the parser does not always require a lookahead token
1486 in order to detect a syntax error. Because no unexpected token or
1487 expected tokens can then be reported, the verbose syntax error
1488 message described above is suppressed, and the parser instead
e4ab1254 1489 reports the simpler message, "syntax error". Previously, this
d2060f06
JD
1490 suppression was sometimes erroneously triggered by %nonassoc when a
1491 lookahead was actually required. Now verbose messages are
1492 suppressed only when all previous lookaheads have already been
1493 shifted or discarded.
1494
1495*** Previously, the list of expected tokens erroneously included tokens
1496 that would actually induce a syntax error because conflicts for them
1497 were resolved with %nonassoc in the current parser state. Such
1498 tokens are now properly omitted from the list.
1499
1500*** Expected token lists are still often wrong due to state merging
fcf834f9
JD
1501 (from LALR or IELR) and default reductions, which can both add
1502 invalid tokens and subtract valid tokens. Canonical LR almost
1503 completely fixes this problem by eliminating state merging and
1504 default reductions. However, there is one minor problem left even
1505 when using canonical LR and even after the fixes above. That is,
1506 if the resolution of a conflict with %nonassoc appears in a later
1507 parser state than the one at which some syntax error is
1508 discovered, the conflicted token is still erroneously included in
1509 the expected token list. Bison's new LAC implementation,
1510 described above, eliminates this problem and the need for
1511 canonical LR. However, LAC is still experimental and is disabled
1512 by default.
53f036ce 1513
1a33f4f6
JD
1514** Java skeleton fixes:
1515
1516*** A location handling bug has been fixed.
1517
1518*** The top element of each of the value stack and location stack is now
1519 cleared when popped so that it can be garbage collected.
6771a463 1520
02803d55
JD
1521*** Parser traces now print the top element of the stack.
1522
86408959
JD
1523** -W/--warnings fixes:
1524
e4ab1254 1525*** Bison now properly recognizes the "no-" versions of categories:
86408959
JD
1526
1527 For example, given the following command line, Bison now enables all
1528 warnings except warnings for incompatibilities with POSIX Yacc:
1529
1530 bison -Wall,no-yacc gram.y
1531
786743d5
JD
1532*** Bison now treats S/R and R/R conflicts like other warnings:
1533
1534 Previously, conflict reports were independent of Bison's normal
1535 warning system. Now, Bison recognizes the warning categories
e4ab1254 1536 "conflicts-sr" and "conflicts-rr". This change has important
786743d5
JD
1537 consequences for the -W and --warnings command-line options. For
1538 example:
1539
1540 bison -Wno-conflicts-sr gram.y # S/R conflicts not reported
1541 bison -Wno-conflicts-rr gram.y # R/R conflicts not reported
1542 bison -Wnone gram.y # no conflicts are reported
1543 bison -Werror gram.y # any conflict is an error
1544
1545 However, as before, if the %expect or %expect-rr directive is
1546 specified, an unexpected number of conflicts is an error, and an
1547 expected number of conflicts is not reported, so -W and --warning
1548 then have no effect on the conflict report.
1549
e4ab1254 1550*** The "none" category no longer disables a preceding "error":
bf0e44e8
JD
1551
1552 For example, for the following command line, Bison now reports
1553 errors instead of warnings for incompatibilities with POSIX Yacc:
1554
1555 bison -Werror,none,yacc gram.y
1556
e4ab1254 1557*** The "none" category now disables all Bison warnings:
c39014ae 1558
e4ab1254 1559 Previously, the "none" category disabled only Bison warnings for
c39014ae
JD
1560 which there existed a specific -W/--warning category. However,
1561 given the following command line, Bison is now guaranteed to
1562 suppress all warnings:
1563
1564 bison -Wnone gram.y
1565
1f36f544
JD
1566** Precedence directives can now assign token number 0:
1567
1568 Since Bison 2.3b, which restored the ability of precedence
1569 directives to assign token numbers, doing so for token number 0 has
1570 produced an assertion failure. For example:
1571
1572 %left END 0
1573
1574 This bug has been fixed.
1575
64877e5e 1576* Changes in version 2.4.3 (2010-08-05):
8b9e021f 1577
2bfcac9a
JD
1578** Bison now obeys -Werror and --warnings=error for warnings about
1579 grammar rules that are useless in the parser due to conflicts.
1580
8b9e021f
JD
1581** Problems with spawning M4 on at least FreeBSD 8 and FreeBSD 9 have
1582 been fixed.
1583
4ad3921d
JD
1584** Failures in the test suite for GCC 4.5 have been fixed.
1585
06cb07d5
JD
1586** Failures in the test suite for some versions of Sun Studio C++ have
1587 been fixed.
1588
9b5049bd
JD
1589** Contrary to Bison 2.4.2's NEWS entry, it has been decided that
1590 warnings about undefined %prec identifiers will not be converted to
1591 errors in Bison 2.5. They will remain warnings, which should be
1592 sufficient for POSIX while avoiding backward compatibility issues.
1593
93d7dde9
JD
1594** Minor documentation fixes.
1595
e19a049c 1596* Changes in version 2.4.2 (2010-03-20):
74553c98 1597
f39ab286
JD
1598** Some portability problems that resulted in failures and livelocks
1599 in the test suite on some versions of at least Solaris, AIX, HP-UX,
e19a049c
JD
1600 RHEL4, and Tru64 have been addressed. As a result, fatal Bison
1601 errors should no longer cause M4 to report a broken pipe on the
f39ab286
JD
1602 affected platforms.
1603
e4ab1254 1604** "%prec IDENTIFIER" requires IDENTIFIER to be defined separately.
8bb3a2e7
JD
1605
1606 POSIX specifies that an error be reported for any identifier that does
1607 not appear on the LHS of a grammar rule and that is not defined by
1608 %token, %left, %right, or %nonassoc. Bison 2.3b and later lost this
1609 error report for the case when an identifier appears only after a
1610 %prec directive. It is now restored. However, for backward
1611 compatibility with recent Bison releases, it is only a warning for
1612 now. In Bison 2.5 and later, it will return to being an error.
9b5049bd
JD
1613 [Between the 2.4.2 and 2.4.3 releases, it was decided that this
1614 warning will not be converted to an error in Bison 2.5.]
8bb3a2e7 1615
d8911864
EB
1616** Detection of GNU M4 1.4.6 or newer during configure is improved.
1617
a603c6e0
JD
1618** Warnings from gcc's -Wundef option about undefined YYENABLE_NLS,
1619 YYLTYPE_IS_TRIVIAL, and __STRICT_ANSI__ in C/C++ parsers are now
1620 avoided.
c938d650 1621
98a345a2
JD
1622** %code is now a permanent feature.
1623
1624 A traditional Yacc prologue directive is written in the form:
1625
1626 %{CODE%}
1627
1628 To provide a more flexible alternative, Bison 2.3b introduced the
1629 %code directive with the following forms for C/C++:
1630
1631 %code {CODE}
1632 %code requires {CODE}
1633 %code provides {CODE}
1634 %code top {CODE}
1635
1636 These forms are now considered permanent features of Bison. See the
1637 %code entries in the section "Bison Declaration Summary" in the Bison
1638 manual for a summary of their functionality. See the section
1639 "Prologue Alternatives" for a detailed discussion including the
1640 advantages of %code over the traditional Yacc prologue directive.
1641
1642 Bison's Java feature as a whole including its current usage of %code
1643 is still considered experimental.
1644
1625df5b
JD
1645** YYFAIL is deprecated and will eventually be removed.
1646
1647 YYFAIL has existed for many years as an undocumented feature of
1648 deterministic parsers in C generated by Bison. Previously, it was
1649 documented for Bison's experimental Java parsers. YYFAIL is no longer
1650 documented for Java parsers and is formally deprecated in both cases.
1651 Users are strongly encouraged to migrate to YYERROR, which is
1652 specified by POSIX.
1653
1654 Like YYERROR, you can invoke YYFAIL from a semantic action in order to
1655 induce a syntax error. The most obvious difference from YYERROR is
1656 that YYFAIL will automatically invoke yyerror to report the syntax
1657 error so that you don't have to. However, there are several other
1658 subtle differences between YYERROR and YYFAIL, and YYFAIL suffers from
e4ab1254 1659 inherent flaws when %error-verbose or "#define YYERROR_VERBOSE" is
1625df5b
JD
1660 used. For a more detailed discussion, see:
1661
1662 http://lists.gnu.org/archive/html/bison-patches/2009-12/msg00024.html
1663
1664 The upcoming Bison 2.5 will remove YYFAIL from Java parsers, but
1665 deterministic parsers in C will continue to implement it. However,
1666 because YYFAIL is already flawed, it seems futile to try to make new
1667 Bison features compatible with it. Thus, during parser generation,
1668 Bison 2.5 will produce a warning whenever it discovers YYFAIL in a
1669 rule action. In a later release, YYFAIL will be disabled for
e4ab1254 1670 %error-verbose and "#define YYERROR_VERBOSE". Eventually, YYFAIL will
1625df5b
JD
1671 be removed altogether.
1672
1673 There exists at least one case where Bison 2.5's YYFAIL warning will
1674 be a false positive. Some projects add phony uses of YYFAIL and other
1675 Bison-defined macros for the sole purpose of suppressing C
1676 preprocessor warnings (from GCC cpp's -Wunused-macros, for example).
1677 To avoid Bison's future warning, such YYFAIL uses can be moved to the
e4ab1254 1678 epilogue (that is, after the second "%%") in the Bison input file. In
1625df5b
JD
1679 this release (2.4.2), Bison already generates its own code to suppress
1680 C preprocessor warnings for YYFAIL, so projects can remove their own
1681 phony uses of YYFAIL if compatibility with Bison releases prior to
1682 2.4.2 is not necessary.
1683
2755de8f
AD
1684** Internationalization.
1685
1686 Fix a regression introduced in Bison 2.4: Under some circumstances,
1687 message translations were not installed although supported by the
1688 host system.
1689
74553c98 1690* Changes in version 2.4.1 (2008-12-11):
c9ba9e59 1691
0ea583d2
AD
1692** In the GLR defines file, unexpanded M4 macros in the yylval and yylloc
1693 declarations have been fixed.
1979121c 1694
0ea583d2
AD
1695** Temporary hack for adding a semicolon to the user action.
1696
1697 Bison used to prepend a trailing semicolon at the end of the user
1698 action for reductions. This allowed actions such as
1699
1700 exp: exp "+" exp { $$ = $1 + $3 };
1701
1702 instead of
1703
1704 exp: exp "+" exp { $$ = $1 + $3; };
1705
e4ab1254 1706 Some grammars still depend on this "feature". Bison 2.4.1 restores
0ea583d2
AD
1707 the previous behavior in the case of C output (specifically, when
1708 neither %language or %skeleton or equivalent command-line options
1709 are used) to leave more time for grammars depending on the old
1710 behavior to be adjusted. Future releases of Bison will disable this
1711 feature.
1712
1713** A few minor improvements to the Bison manual.
c9ba9e59 1714
402b123d 1715* Changes in version 2.4 (2008-11-02):
7bd1665a 1716
402b123d 1717** %language is an experimental feature.
ed4d67dc
JD
1718
1719 We first introduced this feature in test release 2.3b as a cleaner
1720 alternative to %skeleton. Since then, we have discussed the possibility of
1721 modifying its effect on Bison's output file names. Thus, in this release,
1722 we consider %language to be an experimental feature that will likely evolve
1723 in future releases.
7bd1665a 1724
402b123d 1725** Forward compatibility with GNU M4 has been improved.
241fda7a 1726
402b123d 1727** Several bugs in the C++ skeleton and the experimental Java skeleton have been
241fda7a
JD
1728 fixed.
1729
402b123d 1730* Changes in version 2.3b (2008-05-27):
35fe0834 1731
402b123d 1732** The quotes around NAME that used to be required in the following directive
d9df47b6
JD
1733 are now deprecated:
1734
1735 %define NAME "VALUE"
1736
e4ab1254 1737** The directive "%pure-parser" is now deprecated in favor of:
d9df47b6
JD
1738
1739 %define api.pure
1740
1741 which has the same effect except that Bison is more careful to warn about
1742 unreasonable usage in the latter case.
1743
402b123d 1744** Push Parsing
c373bf8b
JD
1745
1746 Bison can now generate an LALR(1) parser in C with a push interface. That
e4ab1254
AD
1747 is, instead of invoking "yyparse", which pulls tokens from "yylex", you can
1748 push one token at a time to the parser using "yypush_parse", which will
c373bf8b
JD
1749 return to the caller after processing each token. By default, the push
1750 interface is disabled. Either of the following directives will enable it:
1751
1752 %define api.push_pull "push" // Just push; does not require yylex.
1753 %define api.push_pull "both" // Push and pull; requires yylex.
1754
e4ab1254 1755 See the new section "A Push Parser" in the Bison manual for details.
c373bf8b 1756
59da312b
JD
1757 The current push parsing interface is experimental and may evolve. More user
1758 feedback will help to stabilize it.
1759
402b123d 1760** The -g and --graph options now output graphs in Graphviz DOT format,
8e55b3aa
JD
1761 not VCG format. Like --graph, -g now also takes an optional FILE argument
1762 and thus cannot be bundled with other short options.
c373bf8b 1763
402b123d 1764** Java
59da312b
JD
1765
1766 Bison can now generate an LALR(1) parser in Java. The skeleton is
e4ab1254 1767 "data/lalr1.java". Consider using the new %language directive instead of
59da312b
JD
1768 %skeleton to select it.
1769
e4ab1254 1770 See the new section "Java Parsers" in the Bison manual for details.
59da312b
JD
1771
1772 The current Java interface is experimental and may evolve. More user
1773 feedback will help to stabilize it.
2bd435c3 1774 Contributed by Paolo Bonzini.
59da312b 1775
402b123d 1776** %language
59da312b
JD
1777
1778 This new directive specifies the programming language of the generated
d43f77e7
PB
1779 parser, which can be C (the default), C++, or Java. Besides the skeleton
1780 that Bison uses, the directive affects the names of the generated files if
1781 the grammar file's name ends in ".y".
59da312b 1782
402b123d 1783** XML Automaton Report
59da312b
JD
1784
1785 Bison can now generate an XML report of the LALR(1) automaton using the new
e4ab1254 1786 "--xml" option. The current XML schema is experimental and may evolve. More
59da312b 1787 user feedback will help to stabilize it.
2bd435c3 1788 Contributed by Wojciech Polak.
c373bf8b 1789
402b123d 1790** The grammar file may now specify the name of the parser header file using
c373bf8b
JD
1791 %defines. For example:
1792
1793 %defines "parser.h"
1794
402b123d 1795** When reporting useless rules, useless nonterminals, and unused terminals,
d80fb37a
JD
1796 Bison now employs the terms "useless in grammar" instead of "useless",
1797 "useless in parser" instead of "never reduced", and "unused in grammar"
1798 instead of "unused".
cff03fb2 1799
402b123d 1800** Unreachable State Removal
c373bf8b
JD
1801
1802 Previously, Bison sometimes generated parser tables containing unreachable
31984206
JD
1803 states. A state can become unreachable during conflict resolution if Bison
1804 disables a shift action leading to it from a predecessor state. Bison now:
75ad86ee
JD
1805
1806 1. Removes unreachable states.
1807
1808 2. Does not report any conflicts that appeared in unreachable states.
1809 WARNING: As a result, you may need to update %expect and %expect-rr
1810 directives in existing grammar files.
1811
1812 3. For any rule used only in such states, Bison now reports the rule as
cff03fb2 1813 "useless in parser due to conflicts".
75ad86ee 1814
31984206
JD
1815 This feature can be disabled with the following directive:
1816
1817 %define lr.keep_unreachable_states
1818
e4ab1254 1819 See the %define entry in the "Bison Declaration Summary" in the Bison manual
31984206
JD
1820 for further discussion.
1821
e4ab1254 1822** Lookahead Set Correction in the ".output" Report
b1cc23c4 1823
e4ab1254
AD
1824 When instructed to generate a ".output" file including lookahead sets
1825 (using "--report=lookahead", for example), Bison now prints each reduction's
88c78747
JD
1826 lookahead set only next to the associated state's one item that (1) is
1827 associated with the same rule as the reduction and (2) has its dot at the end
1828 of its RHS. Previously, Bison also erroneously printed the lookahead set
1829 next to all of the state's other items associated with the same rule. This
e4ab1254 1830 bug affected only the ".output" file and not the generated parser source
88c78747
JD
1831 code.
1832
e4ab1254 1833** --report-file=FILE is a new option to override the default ".output" file
59da312b 1834 name.
1bb2bd75 1835
e4ab1254 1836** The "=" that used to be required in the following directives is now
02975b9a
JD
1837 deprecated:
1838
1839 %file-prefix "parser"
1840 %name-prefix "c_"
1841 %output "parser.c"
1842
e4ab1254 1843** An Alternative to "%{...%}" -- "%code QUALIFIER {CODE}"
c373bf8b
JD
1844
1845 Bison 2.3a provided a new set of directives as a more flexible alternative to
8e0a5e9e
JD
1846 the traditional Yacc prologue blocks. Those have now been consolidated into
1847 a single %code directive with an optional qualifier field, which identifies
1848 the purpose of the code and thus the location(s) where Bison should generate
1849 it:
1850
e4ab1254
AD
1851 1. "%code {CODE}" replaces "%after-header {CODE}"
1852 2. "%code requires {CODE}" replaces "%start-header {CODE}"
1853 3. "%code provides {CODE}" replaces "%end-header {CODE}"
1854 4. "%code top {CODE}" replaces "%before-header {CODE}"
8e0a5e9e 1855
e4ab1254
AD
1856 See the %code entries in section "Bison Declaration Summary" in the Bison
1857 manual for a summary of the new functionality. See the new section "Prologue
1858 Alternatives" for a detailed discussion including the advantages of %code
8e0a5e9e
JD
1859 over the traditional Yacc prologues.
1860
1861 The prologue alternatives are experimental. More user feedback will help to
1862 determine whether they should become permanent features.
1863
402b123d 1864** Revised warning: unset or unused mid-rule values
17bd8a73
JD
1865
1866 Since Bison 2.2, Bison has warned about mid-rule values that are set but not
1867 used within any of the actions of the parent rule. For example, Bison warns
1868 about unused $2 in:
1869
1870 exp: '1' { $$ = 1; } '+' exp { $$ = $1 + $4; };
1871
1872 Now, Bison also warns about mid-rule values that are used but not set. For
1873 example, Bison warns about unset $$ in the mid-rule action in:
1874
1875 exp: '1' { $1 = 1; } '+' exp { $$ = $2 + $4; };
1876
1877 However, Bison now disables both of these warnings by default since they
1878 sometimes prove to be false alarms in existing grammars employing the Yacc
1879 constructs $0 or $-N (where N is some positive integer).
1880
e4ab1254
AD
1881 To enable these warnings, specify the option "--warnings=midrule-values" or
1882 "-W", which is a synonym for "--warnings=all".
17bd8a73 1883
e4ab1254 1884** Default %destructor or %printer with "<*>" or "<>"
c373bf8b
JD
1885
1886 Bison now recognizes two separate kinds of default %destructor's and
12e35840
JD
1887 %printer's:
1888
e4ab1254 1889 1. Place "<*>" in a %destructor/%printer symbol list to define a default
12e35840
JD
1890 %destructor/%printer for all grammar symbols for which you have formally
1891 declared semantic type tags.
1892
e4ab1254 1893 2. Place "<>" in a %destructor/%printer symbol list to define a default
12e35840
JD
1894 %destructor/%printer for all grammar symbols without declared semantic
1895 type tags.
1896
e4ab1254
AD
1897 Bison no longer supports the "%symbol-default" notation from Bison 2.3a.
1898 "<*>" and "<>" combined achieve the same effect with one exception: Bison no
12e35840
JD
1899 longer applies any %destructor to a mid-rule value if that mid-rule value is
1900 not actually ever referenced using either $$ or $n in a semantic action.
1901
85894313
JD
1902 The default %destructor's and %printer's are experimental. More user
1903 feedback will help to determine whether they should become permanent
1904 features.
1905
e4ab1254 1906 See the section "Freeing Discarded Symbols" in the Bison manual for further
12e35840
JD
1907 details.
1908
402b123d 1909** %left, %right, and %nonassoc can now declare token numbers. This is required
e4ab1254 1910 by POSIX. However, see the end of section "Operator Precedence" in the Bison
ab7f29f8
JD
1911 manual for a caveat concerning the treatment of literal strings.
1912
402b123d 1913** The nonfunctional --no-parser, -n, and %no-parser options have been
b1cc23c4
JD
1914 completely removed from Bison.
1915
402b123d 1916* Changes in version 2.3a, 2006-09-13:
742e4900 1917
402b123d 1918** Instead of %union, you can define and use your own union type
ddc8ede1
PE
1919 YYSTYPE if your grammar contains at least one <type> tag.
1920 Your YYSTYPE need not be a macro; it can be a typedef.
1921 This change is for compatibility with other Yacc implementations,
1922 and is required by POSIX.
1923
402b123d 1924** Locations columns and lines start at 1.
cd48d21d
AD
1925 In accordance with the GNU Coding Standards and Emacs.
1926
402b123d 1927** You may now declare per-type and default %destructor's and %printer's:
ec5479ce
JD
1928
1929 For example:
1930
b2a0b7ca
JD
1931 %union { char *string; }
1932 %token <string> STRING1
1933 %token <string> STRING2
1934 %type <string> string1
1935 %type <string> string2
1936 %union { char character; }
1937 %token <character> CHR
1938 %type <character> chr
1939 %destructor { free ($$); } %symbol-default
1940 %destructor { free ($$); printf ("%d", @$.first_line); } STRING1 string1
1941 %destructor { } <character>
1942
1943 guarantees that, when the parser discards any user-defined symbol that has a
e4ab1254
AD
1944 semantic type tag other than "<character>", it passes its semantic value to
1945 "free". However, when the parser discards a "STRING1" or a "string1", it
1946 also prints its line number to "stdout". It performs only the second
1947 "%destructor" in this case, so it invokes "free" only once.
ec5479ce 1948
85894313
JD
1949 [Although we failed to mention this here in the 2.3a release, the default
1950 %destructor's and %printer's were experimental, and they were rewritten in
1951 future versions.]
1952
e4ab1254
AD
1953** Except for LALR(1) parsers in C with POSIX Yacc emulation enabled (with "-y",
1954 "--yacc", or "%yacc"), Bison no longer generates #define statements for
b931235e
JD
1955 associating token numbers with token names. Removing the #define statements
1956 helps to sanitize the global namespace during preprocessing, but POSIX Yacc
1957 requires them. Bison still generates an enum for token names in all cases.
1958
402b123d 1959** Handling of traditional Yacc prologue blocks is now more consistent but
34f98f46 1960 potentially incompatible with previous releases of Bison.
9bc0dd67
JD
1961
1962 As before, you declare prologue blocks in your grammar file with the
e4ab1254 1963 "%{ ... %}" syntax. To generate the pre-prologue, Bison concatenates all
34f98f46
JD
1964 prologue blocks that you've declared before the first %union. To generate
1965 the post-prologue, Bison concatenates all prologue blocks that you've
ddc8ede1 1966 declared after the first %union.
9bc0dd67 1967
34f98f46 1968 Previous releases of Bison inserted the pre-prologue into both the header
9bc0dd67
JD
1969 file and the code file in all cases except for LALR(1) parsers in C. In the
1970 latter case, Bison inserted it only into the code file. For parsers in C++,
1971 the point of insertion was before any token definitions (which associate
1972 token numbers with names). For parsers in C, the point of insertion was
1973 after the token definitions.
1974
1975 Now, Bison never inserts the pre-prologue into the header file. In the code
1976 file, it always inserts it before the token definitions.
1977
402b123d 1978** Bison now provides a more flexible alternative to the traditional Yacc
34f98f46
JD
1979 prologue blocks: %before-header, %start-header, %end-header, and
1980 %after-header.
1981
1982 For example, the following declaration order in the grammar file reflects the
1983 order in which Bison will output these code blocks. However, you are free to
1984 declare these code blocks in your grammar file in whatever order is most
1985 convenient for you:
1986
1987 %before-header {
1988 /* Bison treats this block like a pre-prologue block: it inserts it into
1989 * the code file before the contents of the header file. It does *not*
1990 * insert it into the header file. This is a good place to put
1991 * #include's that you want at the top of your code file. A common
e4ab1254 1992 * example is '#include "system.h"'. */
34f98f46
JD
1993 }
1994 %start-header {
1995 /* Bison inserts this block into both the header file and the code file.
1996 * In both files, the point of insertion is before any Bison-generated
1997 * token, semantic type, location type, and class definitions. This is a
1998 * good place to define %union dependencies, for example. */
9bc0dd67
JD
1999 }
2000 %union {
34f98f46
JD
2001 /* Unlike the traditional Yacc prologue blocks, the output order for the
2002 * new %*-header blocks is not affected by their declaration position
2003 * relative to any %union in the grammar file. */
9bc0dd67 2004 }
34f98f46
JD
2005 %end-header {
2006 /* Bison inserts this block into both the header file and the code file.
2007 * In both files, the point of insertion is after the Bison-generated
2008 * definitions. This is a good place to declare or define public
2009 * functions or data structures that depend on the Bison-generated
2010 * definitions. */
9bc0dd67 2011 }
34f98f46
JD
2012 %after-header {
2013 /* Bison treats this block like a post-prologue block: it inserts it into
2014 * the code file after the contents of the header file. It does *not*
2015 * insert it into the header file. This is a good place to declare or
2016 * define internal functions or data structures that depend on the
2017 * Bison-generated definitions. */
2018 }
2019
2020 If you have multiple occurrences of any one of the above declarations, Bison
2021 will concatenate the contents in declaration order.
9bc0dd67 2022
85894313
JD
2023 [Although we failed to mention this here in the 2.3a release, the prologue
2024 alternatives were experimental, and they were rewritten in future versions.]
2025
e4ab1254 2026** The option "--report=look-ahead" has been changed to "--report=lookahead".
9e6e7ed2
PE
2027 The old spelling still works, but is not documented and may be removed
2028 in a future release.
742e4900 2029
402b123d 2030* Changes in version 2.3, 2006-06-05:
4ad3ed84 2031
e4ab1254 2032** GLR grammars should now use "YYRECOVERING ()" instead of "YYRECOVERING",
4ad3ed84
PE
2033 for compatibility with LALR(1) grammars.
2034
402b123d 2035** It is now documented that any definition of YYSTYPE or YYLTYPE should
4ad3ed84
PE
2036 be to a type name that does not contain parentheses or brackets.
2037
402b123d 2038* Changes in version 2.2, 2006-05-19:
193d7c70 2039
402b123d 2040** The distribution terms for all Bison-generated parsers now permit
193d7c70
PE
2041 using the parsers in nonfree programs. Previously, this permission
2042 was granted only for Bison-generated LALR(1) parsers in C.
5f4236a0 2043
402b123d 2044** %name-prefix changes the namespace name in C++ outputs.
aa08666d 2045
402b123d 2046** The C++ parsers export their token_type.
5f4236a0 2047
402b123d 2048** Bison now allows multiple %union declarations, and concatenates
d6ca7905
PE
2049 their contents together.
2050
402b123d 2051** New warning: unused values
4d7bc38c
PE
2052 Right-hand side symbols whose values are not used are reported,
2053 if the symbols have destructors. For instance:
affac613 2054
8f3596a6 2055 exp: exp "?" exp ":" exp { $1 ? $1 : $3; }
e9690142
JD
2056 | exp "+" exp
2057 ;
affac613 2058
8f3596a6
AD
2059 will trigger a warning about $$ and $5 in the first rule, and $3 in
2060 the second ($1 is copied to $$ by the default rule). This example
4e26c69e 2061 most likely contains three errors, and could be rewritten as:
affac613 2062
4e26c69e 2063 exp: exp "?" exp ":" exp
e9690142
JD
2064 { $$ = $1 ? $3 : $5; free ($1 ? $5 : $3); free ($1); }
2065 | exp "+" exp
2066 { $$ = $1 ? $1 : $3; if ($1) free ($3); }
2067 ;
affac613 2068
4e26c69e
PE
2069 However, if the original actions were really intended, memory leaks
2070 and all, the warnings can be suppressed by letting Bison believe the
2071 values are used, e.g.:
721be13c 2072
8f3596a6 2073 exp: exp "?" exp ":" exp { $1 ? $1 : $3; (void) ($$, $5); }
e9690142
JD
2074 | exp "+" exp { $$ = $1; (void) $3; }
2075 ;
721be13c 2076
84866159
AD
2077 If there are mid-rule actions, the warning is issued if no action
2078 uses it. The following triggers no warning: $1 and $3 are used.
2079
2080 exp: exp { push ($1); } '+' exp { push ($3); sum (); };
2081
721be13c
PE
2082 The warning is intended to help catching lost values and memory leaks.
2083 If a value is ignored, its associated memory typically is not reclaimed.
affac613 2084
402b123d 2085** %destructor vs. YYABORT, YYACCEPT, and YYERROR.
9d9b8b70
PE
2086 Destructors are now called when user code invokes YYABORT, YYACCEPT,
2087 and YYERROR, for all objects on the stack, other than objects
2088 corresponding to the right-hand side of the current rule.
a85284cf 2089
402b123d 2090** %expect, %expect-rr
035aa4a0
PE
2091 Incorrect numbers of expected conflicts are now actual errors,
2092 instead of warnings.
2093
402b123d 2094** GLR, YACC parsers.
4e26c69e
PE
2095 The %parse-params are available in the destructors (and the
2096 experimental printers) as per the documentation.
4b367315 2097
e4ab1254 2098** Bison now warns if it finds a stray "$" or "@" in an action.
ad6a9b97 2099
402b123d 2100** %require "VERSION"
4e26c69e
PE
2101 This specifies that the grammar file depends on features implemented
2102 in Bison version VERSION or higher.
b50d2359 2103
402b123d 2104** lalr1.cc: The token and value types are now class members.
e14d0ab6
AD
2105 The tokens were defined as free form enums and cpp macros. YYSTYPE
2106 was defined as a free form union. They are now class members:
e4ab1254
AD
2107 tokens are enumerations of the "yy::parser::token" struct, and the
2108 semantic values have the "yy::parser::semantic_type" type.
fb9712a9
AD
2109
2110 If you do not want or can update to this scheme, the directive
e4ab1254 2111 '%define "global_tokens_and_yystype" "1"' triggers the global
b50d2359
AD
2112 definition of tokens and YYSTYPE. This change is suitable both
2113 for previous releases of Bison, and this one.
fb9712a9 2114
b50d2359 2115 If you wish to update, then make sure older version of Bison will
e4ab1254 2116 fail using '%require "2.2"'.
fb9712a9 2117
402b123d 2118** DJGPP support added.
193d7c70 2119\f
402b123d 2120* Changes in version 2.1, 2005-09-16:
1ce59070 2121
402b123d 2122** The C++ lalr1.cc skeleton supports %lex-param.
e14d0ab6 2123
402b123d 2124** Bison-generated parsers now support the translation of diagnostics like
baf785db
PE
2125 "syntax error" into languages other than English. The default
2126 language is still English. For details, please see the new
0410a6e0
PE
2127 Internationalization section of the Bison manual. Software
2128 distributors should also see the new PACKAGING file. Thanks to
2129 Bruno Haible for this new feature.
1ce59070 2130
402b123d 2131** Wording in the Bison-generated parsers has been changed slightly to
1a059451
PE
2132 simplify translation. In particular, the message "memory exhausted"
2133 has replaced "parser stack overflow", as the old message was not
2134 always accurate for modern Bison-generated parsers.
2135
402b123d 2136** Destructors are now called when the parser aborts, for all symbols left
258b75ca
PE
2137 behind on the stack. Also, the start symbol is now destroyed after a
2138 successful parse. In both cases, the behavior was formerly inconsistent.
2139
402b123d 2140** When generating verbose diagnostics, Bison-generated parsers no longer
72f000b0
PE
2141 quote the literal strings associated with tokens. For example, for
2142 a syntax error associated with '%token NUM "number"' they might
2143 print 'syntax error, unexpected number' instead of 'syntax error,
2144 unexpected "number"'.
193d7c70 2145\f
402b123d 2146* Changes in version 2.0, 2004-12-25:
efeed023 2147
402b123d 2148** Possibly-incompatible changes
d7e14fc0 2149
82de6b0d
PE
2150 - Bison-generated parsers no longer default to using the alloca function
2151 (when available) to extend the parser stack, due to widespread
2152 problems in unchecked stack-overflow detection. You can "#define
2153 YYSTACK_USE_ALLOCA 1" to require the use of alloca, but please read
2154 the manual to determine safe values for YYMAXDEPTH in that case.
8dd162d3 2155
82de6b0d
PE
2156 - Error token location.
2157 During error recovery, the location of the syntax error is updated
2158 to cover the whole sequence covered by the error token: it includes
2159 the shifted symbols thrown away during the first part of the error
2160 recovery, and the lookahead rejected during the second part.
18d192f0 2161
82de6b0d
PE
2162 - Semicolon changes:
2163 . Stray semicolons are no longer allowed at the start of a grammar.
2164 . Semicolons are now required after in-grammar declarations.
e342c3be 2165
82de6b0d
PE
2166 - Unescaped newlines are no longer allowed in character constants or
2167 string literals. They were never portable, and GCC 3.4.0 has
2168 dropped support for them. Better diagnostics are now generated if
2169 forget a closing quote.
8dd162d3 2170
82de6b0d 2171 - NUL bytes are no longer allowed in Bison string literals, unfortunately.
f74b6f91 2172
402b123d 2173** New features
1452af69 2174
82de6b0d 2175 - GLR grammars now support locations.
4febdd96 2176
82de6b0d
PE
2177 - New directive: %initial-action.
2178 This directive allows the user to run arbitrary code (including
2179 initializing @$) from yyparse before parsing starts.
1452af69 2180
82de6b0d
PE
2181 - A new directive "%expect-rr N" specifies the expected number of
2182 reduce/reduce conflicts in GLR parsers.
1452af69 2183
e4ab1254 2184 - %token numbers can now be hexadecimal integers, e.g., "%token FOO 0x12d".
82de6b0d 2185 This is a GNU extension.
4febdd96 2186
e4ab1254 2187 - The option "--report=lookahead" was changed to "--report=look-ahead".
9e6e7ed2 2188 [However, this was changed back after 2.3.]
1452af69 2189
82de6b0d 2190 - Experimental %destructor support has been added to lalr1.cc.
1452af69 2191
82de6b0d
PE
2192 - New configure option --disable-yacc, to disable installation of the
2193 yacc command and -ly library introduced in 1.875 for POSIX conformance.
6040d338 2194
402b123d 2195** Bug fixes
d5a3fe37 2196
82de6b0d
PE
2197 - For now, %expect-count violations are now just warnings, not errors.
2198 This is for compatibility with Bison 1.75 and earlier (when there are
2199 reduce/reduce conflicts) and with Bison 1.30 and earlier (when there
2200 are too many or too few shift/reduce conflicts). However, in future
2201 versions of Bison we plan to improve the %expect machinery so that
2202 these violations will become errors again.
3473d0f8 2203
82de6b0d
PE
2204 - Within Bison itself, numbers (e.g., goto numbers) are no longer
2205 arbitrarily limited to 16-bit counts.
d600ee67 2206
82de6b0d 2207 - Semicolons are now allowed before "|" in grammar rules, as POSIX requires.
d600ee67 2208\f
402b123d 2209* Changes in version 1.875, 2003-01-01:
963fcc17 2210
402b123d 2211** The documentation license has been upgraded to version 1.2
dc546b0f 2212 of the GNU Free Documentation License.
75eb3bc4 2213
402b123d 2214** syntax error processing
75eb3bc4 2215
dc546b0f
PE
2216 - In Yacc-style parsers YYLLOC_DEFAULT is now used to compute error
2217 locations too. This fixes bugs in error-location computation.
75eb3bc4 2218
dc546b0f
PE
2219 - %destructor
2220 It is now possible to reclaim the memory associated to symbols
2221 discarded during error recovery. This feature is still experimental.
20daca06 2222
dc546b0f
PE
2223 - %error-verbose
2224 This new directive is preferred over YYERROR_VERBOSE.
74724a70 2225
dc546b0f
PE
2226 - #defining yyerror to steal internal variables is discouraged.
2227 It is not guaranteed to work forever.
d1de5372 2228
402b123d 2229** POSIX conformance
d1de5372 2230
dc546b0f
PE
2231 - Semicolons are once again optional at the end of grammar rules.
2232 This reverts to the behavior of Bison 1.33 and earlier, and improves
2233 compatibility with Yacc.
74724a70 2234
e4ab1254
AD
2235 - "parse error" -> "syntax error"
2236 Bison now uniformly uses the term "syntax error"; formerly, the code
2237 and manual sometimes used the term "parse error" instead. POSIX
2238 requires "syntax error" in diagnostics, and it was thought better to
dc546b0f 2239 be consistent.
74724a70 2240
dc546b0f
PE
2241 - The documentation now emphasizes that yylex and yyerror must be
2242 declared before use. C99 requires this.
d1de5372 2243
dc546b0f
PE
2244 - Bison now parses C99 lexical constructs like UCNs and
2245 backslash-newline within C escape sequences, as POSIX 1003.1-2001 requires.
d1de5372 2246
dc546b0f
PE
2247 - File names are properly escaped in C output. E.g., foo\bar.y is
2248 output as "foo\\bar.y".
6780ca7a 2249
dc546b0f 2250 - Yacc command and library now available
e4ab1254 2251 The Bison distribution now installs a "yacc" command, as POSIX requires.
dc546b0f
PE
2252 Also, Bison now installs a small library liby.a containing
2253 implementations of Yacc-compatible yyerror and main functions.
2254 This library is normally not useful, but POSIX requires it.
6e649e65 2255
dc546b0f 2256 - Type clashes now generate warnings, not errors.
6e649e65 2257
dc546b0f
PE
2258 - If the user does not define YYSTYPE as a macro, Bison now declares it
2259 using typedef instead of defining it as a macro.
2260 For consistency, YYLTYPE is also declared instead of defined.
9501dc6e 2261
402b123d 2262** Other compatibility issues
886a425c 2263
e4ab1254
AD
2264 - %union directives can now have a tag before the "{", e.g., the
2265 directive "%union foo {...}" now generates the C code
2266 "typedef union foo { ... } YYSTYPE;"; this is for Yacc compatibility.
2267 The default union tag is "YYSTYPE", for compatibility with Solaris 9 Yacc.
2268 For consistency, YYLTYPE's struct tag is now "YYLTYPE" not "yyltype".
dc546b0f 2269 This is for compatibility with both Yacc and Bison 1.35.
72f889cc 2270
e4ab1254 2271 - ";" is output before the terminating "}" of an action, for
dc546b0f 2272 compatibility with Bison 1.35.
886a425c 2273
dc546b0f 2274 - Bison now uses a Yacc-style format for conflict reports, e.g.,
e4ab1254 2275 "conflicts: 2 shift/reduce, 1 reduce/reduce".
437c2d80 2276
e4ab1254 2277 - "yystype" and "yyltype" are now obsolescent macros instead of being
dc546b0f
PE
2278 typedefs or tags; they are no longer documented and are planned to be
2279 withdrawn in a future release.
2a8d363a 2280
402b123d 2281** GLR parser notes
2a8d363a 2282
dc546b0f
PE
2283 - GLR and inline
2284 Users of Bison have to decide how they handle the portability of the
e4ab1254 2285 C keyword "inline".
959e5f51 2286
e4ab1254
AD
2287 - "parsing stack overflow..." -> "parser stack overflow"
2288 GLR parsers now report "parser stack overflow" as per the Bison manual.
900c5db5 2289
18ad57b3
AD
2290** %parse-param and %lex-param
2291 The macros YYPARSE_PARAM and YYLEX_PARAM provide a means to pass
2292 additional context to yyparse and yylex. They suffer from several
2293 shortcomings:
2294
2295 - a single argument only can be added,
2296 - their types are weak (void *),
242cc08e 2297 - this context is not passed to ancillary functions such as yyerror,
18ad57b3
AD
2298 - only yacc.c parsers support them.
2299
2300 The new %parse-param/%lex-param directives provide a more precise control.
2301 For instance:
2302
2303 %parse-param {int *nastiness}
2304 %lex-param {int *nastiness}
2305 %parse-param {int *randomness}
2306
2307 results in the following signatures:
2308
2309 int yylex (int *nastiness);
2310 int yyparse (int *nastiness, int *randomness);
2311
2312 or, if both %pure-parser and %locations are used:
2313
2314 int yylex (YYSTYPE *lvalp, YYLTYPE *llocp, int *nastiness);
2315 int yyparse (int *nastiness, int *randomness);
2316
402b123d 2317** Bison now warns if it detects conflicting outputs to the same file,
e4ab1254 2318 e.g., it generates a warning for "bison -d -o foo.h foo.y" since
dc546b0f 2319 that command outputs both code and header to foo.h.
6e40b4eb 2320
402b123d 2321** #line in output files
dc546b0f 2322 - --no-line works properly.
6e40b4eb 2323
402b123d 2324** Bison can no longer be built by a K&R C compiler; it requires C89 or
6e40b4eb
AD
2325 later to be built. This change originally took place a few versions
2326 ago, but nobody noticed until we recently asked someone to try
2327 building Bison with a K&R C compiler.
d600ee67 2328\f
402b123d 2329* Changes in version 1.75, 2002-10-14:
7933f2b5 2330
402b123d 2331** Bison should now work on 64-bit hosts.
7933f2b5 2332
402b123d 2333** Indonesian translation thanks to Tedi Heriyanto.
7933f2b5 2334
402b123d 2335** GLR parsers
f50adbbd
AD
2336 Fix spurious parse errors.
2337
402b123d 2338** Pure parsers
f50adbbd
AD
2339 Some people redefine yyerror to steal yyparse' private variables.
2340 Reenable this trick until an official feature replaces it.
2341
402b123d 2342** Type Clashes
d90c934c
AD
2343 In agreement with POSIX and with other Yaccs, leaving a default
2344 action is valid when $$ is untyped, and $1 typed:
2345
e9690142 2346 untyped: ... typed;
d90c934c
AD
2347
2348 but the converse remains an error:
2349
e9690142 2350 typed: ... untyped;
d90c934c 2351
402b123d 2352** Values of mid-rule actions
d90c934c
AD
2353 The following code:
2354
e9690142 2355 foo: { ... } { $$ = $1; } ...
d90c934c
AD
2356
2357 was incorrectly rejected: $1 is defined in the second mid-rule
2358 action, and is equal to the $$ of the first mid-rule action.
d600ee67 2359\f
402b123d 2360* Changes in version 1.50, 2002-10-04:
adc8c848 2361
402b123d 2362** GLR parsing
676385e2
PH
2363 The declaration
2364 %glr-parser
2365 causes Bison to produce a Generalized LR (GLR) parser, capable of handling
2366 almost any context-free grammar, ambiguous or not. The new declarations
e8832397 2367 %dprec and %merge on grammar rules allow parse-time resolution of
676385e2
PH
2368 ambiguities. Contributed by Paul Hilfinger.
2369
7933f2b5 2370 Unfortunately Bison 1.50 does not work properly on 64-bit hosts
420f93c8
PE
2371 like the Alpha, so please stick to 32-bit hosts for now.
2372
402b123d 2373** Output Directory
8c165d89 2374 When not in Yacc compatibility mode, when the output file was not
e4ab1254
AD
2375 specified, running "bison foo/bar.y" created "foo/bar.c". It
2376 now creates "bar.c".
8c165d89 2377
402b123d 2378** Undefined token
007a50a4 2379 The undefined token was systematically mapped to 2 which prevented
e88dbdbf 2380 the use of 2 by the user. This is no longer the case.
007a50a4 2381
402b123d 2382** Unknown token numbers
e88dbdbf 2383 If yylex returned an out of range value, yyparse could die. This is
007a50a4
AD
2384 no longer the case.
2385
402b123d 2386** Error token
e88dbdbf 2387 According to POSIX, the error token must be 256.
23c5a174
AD
2388 Bison extends this requirement by making it a preference: *if* the
2389 user specified that one of her tokens is numbered 256, then error
2390 will be mapped onto another number.
2391
402b123d 2392** Verbose error messages
e4ab1254 2393 They no longer report "..., expecting error or..." for states where
217598da
AD
2394 error recovery is possible.
2395
402b123d 2396** End token
e4ab1254 2397 Defaults to "$end" instead of "$".
217598da 2398
402b123d 2399** Error recovery now conforms to documentation and to POSIX
68cd8af3
PE
2400 When a Bison-generated parser encounters a syntax error, it now pops
2401 the stack until it finds a state that allows shifting the error
2402 token. Formerly, it popped the stack until it found a state that
2403 allowed some non-error action other than a default reduction on the
2404 error token. The new behavior has long been the documented behavior,
2405 and has long been required by POSIX. For more details, please see
337116ba
PE
2406 Paul Eggert, "Reductions during Bison error handling" (2002-05-20)
2407 <http://lists.gnu.org/archive/html/bug-bison/2002-05/msg00038.html>.
68cd8af3 2408
402b123d 2409** Traces
5504898e
AD
2410 Popped tokens and nonterminals are now reported.
2411
402b123d 2412** Larger grammars
a861a339
PE
2413 Larger grammars are now supported (larger token numbers, larger grammar
2414 size (= sum of the LHS and RHS lengths), larger LALR tables).
2415 Formerly, many of these numbers ran afoul of 16-bit limits;
2416 now these limits are 32 bits on most hosts.
355e7c1c 2417
402b123d 2418** Explicit initial rule
643a5994
AD
2419 Bison used to play hacks with the initial rule, which the user does
2420 not write. It is now explicit, and visible in the reports and
2421 graphs as rule 0.
23c5a174 2422
402b123d 2423** Useless rules
643a5994 2424 Before, Bison reported the useless rules, but, although not used,
77714df2 2425 included them in the parsers. They are now actually removed.
23c5a174 2426
402b123d 2427** Useless rules, useless nonterminals
6b98e4b5
AD
2428 They are now reported, as a warning, with their locations.
2429
402b123d 2430** Rules never reduced
e8832397
AD
2431 Rules that can never be reduced because of conflicts are now
2432 reported.
2433
e4ab1254 2434** Incorrect "Token not used"
11652ab3
AD
2435 On a grammar such as
2436
e29f0771
AD
2437 %token useless useful
2438 %%
2439 exp: '0' %prec useful;
11652ab3
AD
2440
2441 where a token was used to set the precedence of the last rule,
e4ab1254 2442 bison reported both "useful" and "useless" as useless tokens.
11652ab3 2443
402b123d 2444** Revert the C++ namespace changes introduced in 1.31
77714df2 2445 as they caused too many portability hassles.
0179dd65 2446
402b123d 2447** Default locations
b2d52318
AD
2448 By an accident of design, the default computation of @$ was
2449 performed after another default computation was performed: @$ = @1.
2450 The latter is now removed: YYLLOC_DEFAULT is fully responsible of
2451 the computation of @$.
adc8c848 2452
402b123d 2453** Token end-of-file
b7c49edf
AD
2454 The token end of file may be specified by the user, in which case,
2455 the user symbol is used in the reports, the graphs, and the verbose
e4ab1254 2456 error messages instead of "$end", which remains being the default.
b7c49edf 2457 For instance
e29f0771 2458 %token MYEOF 0
b7c49edf 2459 or
e29f0771 2460 %token MYEOF 0 "end of file"
fdbcd8e2 2461
402b123d 2462** Semantic parser
fdbcd8e2
AD
2463 This old option, which has been broken for ages, is removed.
2464
402b123d 2465** New translations
a861a339 2466 Brazilian Portuguese, thanks to Alexandre Folle de Menezes.
84614e13
AD
2467 Croatian, thanks to Denis Lackovic.
2468
402b123d 2469** Incorrect token definitions
e4ab1254
AD
2470 When given
2471 %token 'a' "A"
2472 bison used to output
2473 #define 'a' 65
b87f8b21 2474
402b123d 2475** Token definitions as enums
77714df2
AD
2476 Tokens are output both as the traditional #define's, and, provided
2477 the compiler supports ANSI C or is a C++ compiler, as enums.
e88dbdbf 2478 This lets debuggers display names instead of integers.
77714df2 2479
402b123d 2480** Reports
ec3bc396
AD
2481 In addition to --verbose, bison supports --report=THINGS, which
2482 produces additional information:
b408954b
AD
2483 - itemset
2484 complete the core item sets with their closure
e4ab1254 2485 - lookahead [changed to "look-ahead" in 1.875e through 2.3, but changed back]
9e6e7ed2 2486 explicitly associate lookahead tokens to items
b408954b
AD
2487 - solved
2488 describe shift/reduce conflicts solving.
2489 Bison used to systematically output this information on top of
2490 the report. Solved conflicts are now attached to their states.
ec3bc396 2491
402b123d 2492** Type clashes
9af3fbce
AD
2493 Previous versions don't complain when there is a type clash on
2494 the default action if the rule has a mid-rule action, such as in:
2495
e29f0771
AD
2496 %type <foo> bar
2497 %%
2498 bar: '0' {} '0';
9af3fbce
AD
2499
2500 This is fixed.
a861a339 2501
402b123d 2502** GNU M4 is now required when using Bison.
f987e9d2 2503\f
402b123d 2504* Changes in version 1.35, 2002-03-25:
76551463 2505
402b123d 2506** C Skeleton
76551463
AD
2507 Some projects use Bison's C parser with C++ compilers, and define
2508 YYSTYPE as a class. The recent adjustment of C parsers for data
2509 alignment and 64 bit architectures made this impossible.
2510
2511 Because for the time being no real solution for C++ parser
2512 generation exists, kludges were implemented in the parser to
2513 maintain this use. In the future, when Bison has C++ parsers, this
2514 kludge will be disabled.
2515
2516 This kludge also addresses some C++ problems when the stack was
2517 extended.
76551463 2518\f
402b123d 2519* Changes in version 1.34, 2002-03-12:
76551463 2520
402b123d 2521** File name clashes are detected
76551463 2522 $ bison foo.y -d -o foo.x
e4ab1254 2523 fatal error: header and parser would both be named "foo.x"
76551463 2524
e4ab1254 2525** A missing ";" at the end of a rule triggers a warning
76551463
AD
2526 In accordance with POSIX, and in agreement with other
2527 Yacc implementations, Bison will mandate this semicolon in the near
2528 future. This eases the implementation of a Bison parser of Bison
2529 grammars by making this grammar LALR(1) instead of LR(2). To
2530 facilitate the transition, this release introduces a warning.
2531
402b123d 2532** Revert the C++ namespace changes introduced in 1.31, as they caused too
76551463
AD
2533 many portability hassles.
2534
402b123d 2535** DJGPP support added.
76551463 2536
402b123d 2537** Fix test suite portability problems.
76551463 2538\f
402b123d 2539* Changes in version 1.33, 2002-02-07:
76551463 2540
402b123d 2541** Fix C++ issues
76551463
AD
2542 Groff could not be compiled for the definition of size_t was lacking
2543 under some conditions.
2544
402b123d 2545** Catch invalid @n
76551463
AD
2546 As is done with $n.
2547\f
402b123d 2548* Changes in version 1.32, 2002-01-23:
76551463 2549
402b123d 2550** Fix Yacc output file names
76551463 2551
402b123d 2552** Portability fixes
76551463 2553
402b123d 2554** Italian, Dutch translations
76551463 2555\f
402b123d 2556* Changes in version 1.31, 2002-01-14:
52d1aeee 2557
402b123d 2558** Many Bug Fixes
52d1aeee 2559
402b123d 2560** GNU Gettext and %expect
52d1aeee
MA
2561 GNU Gettext asserts 10 s/r conflicts, but there are 7. Now that
2562 Bison dies on incorrect %expectations, we fear there will be
2563 too many bug reports for Gettext, so _for the time being_, %expect
e4ab1254 2564 does not trigger an error when the input file is named "plural.y".
52d1aeee 2565
402b123d 2566** Use of alloca in parsers
52d1aeee
MA
2567 If YYSTACK_USE_ALLOCA is defined to 0, then the parsers will use
2568 malloc exclusively. Since 1.29, but was not NEWS'ed.
2569
2570 alloca is used only when compiled with GCC, to avoid portability
2571 problems as on AIX.
2572
402b123d 2573** yyparse now returns 2 if memory is exhausted; formerly it dumped core.
b47dbebe 2574
402b123d 2575** When the generated parser lacks debugging code, YYDEBUG is now 0
52d1aeee
MA
2576 (as POSIX requires) instead of being undefined.
2577
402b123d 2578** User Actions
52d1aeee
MA
2579 Bison has always permitted actions such as { $$ = $1 }: it adds the
2580 ending semicolon. Now if in Yacc compatibility mode, the semicolon
2581 is no longer output: one has to write { $$ = $1; }.
2582
402b123d 2583** Better C++ compliance
52d1aeee 2584 The output parsers try to respect C++ namespaces.
76551463 2585 [This turned out to be a failed experiment, and it was reverted later.]
52d1aeee 2586
402b123d 2587** Reduced Grammars
52d1aeee
MA
2588 Fixed bugs when reporting useless nonterminals.
2589
402b123d 2590** 64 bit hosts
52d1aeee
MA
2591 The parsers work properly on 64 bit hosts.
2592
402b123d 2593** Error messages
52d1aeee
MA
2594 Some calls to strerror resulted in scrambled or missing error messages.
2595
402b123d 2596** %expect
52d1aeee
MA
2597 When the number of shift/reduce conflicts is correct, don't issue
2598 any warning.
2599
402b123d 2600** The verbose report includes the rule line numbers.
52d1aeee 2601
402b123d 2602** Rule line numbers are fixed in traces.
52d1aeee 2603
402b123d 2604** Swedish translation
52d1aeee 2605
402b123d 2606** Parse errors
52d1aeee
MA
2607 Verbose parse error messages from the parsers are better looking.
2608 Before: parse error: unexpected `'/'', expecting `"number"' or `'-'' or `'(''
2609 Now: parse error: unexpected '/', expecting "number" or '-' or '('
2610
402b123d 2611** Fixed parser memory leaks.
52d1aeee
MA
2612 When the generated parser was using malloc to extend its stacks, the
2613 previous allocations were not freed.
2614
402b123d 2615** Fixed verbose output file.
52d1aeee
MA
2616 Some newlines were missing.
2617 Some conflicts in state descriptions were missing.
2618
402b123d 2619** Fixed conflict report.
52d1aeee
MA
2620 Option -v was needed to get the result.
2621
402b123d 2622** %expect
52d1aeee
MA
2623 Was not used.
2624 Mismatches are errors, not warnings.
2625
402b123d 2626** Fixed incorrect processing of some invalid input.
52d1aeee 2627
402b123d 2628** Fixed CPP guards: 9foo.h uses BISON_9FOO_H instead of 9FOO_H.
52d1aeee 2629
402b123d 2630** Fixed some typos in the documentation.
52d1aeee 2631
402b123d 2632** %token MY_EOF 0 is supported.
52d1aeee
MA
2633 Before, MY_EOF was silently renumbered as 257.
2634
402b123d 2635** doc/refcard.tex is updated.
52d1aeee 2636
402b123d 2637** %output, %file-prefix, %name-prefix.
52d1aeee
MA
2638 New.
2639
402b123d 2640** --output
e4ab1254 2641 New, aliasing "--output-file".
52d1aeee 2642\f
402b123d 2643* Changes in version 1.30, 2001-10-26:
342b8b6e 2644
e4ab1254
AD
2645** "--defines" and "--graph" have now an optional argument which is the
2646 output file name. "-d" and "-g" do not change; they do not take any
342b8b6e
AD
2647 argument.
2648
e4ab1254 2649** "%source_extension" and "%header_extension" are removed, failed
342b8b6e
AD
2650 experiment.
2651
402b123d 2652** Portability fixes.
f987e9d2 2653\f
402b123d 2654* Changes in version 1.29, 2001-09-07:
342b8b6e 2655
402b123d 2656** The output file does not define const, as this caused problems when used
342b8b6e
AD
2657 with common autoconfiguration schemes. If you still use ancient compilers
2658 that lack const, compile with the equivalent of the C compiler option
e4ab1254 2659 "-Dconst=". Autoconf's AC_C_CONST macro provides one way to do this.
342b8b6e 2660
e4ab1254 2661** Added "-g" and "--graph".
f87a2205 2662
402b123d 2663** The Bison manual is now distributed under the terms of the GNU FDL.
f2b5126e 2664
402b123d 2665** The input and the output files has automatically a similar extension.
234a3be3 2666
402b123d 2667** Russian translation added.
f87a2205 2668
402b123d 2669** NLS support updated; should hopefully be less troublesome.
f87a2205 2670
402b123d 2671** Added the old Bison reference card.
c33638bb 2672
e4ab1254 2673** Added "--locations" and "%locations".
6deb4447 2674
e4ab1254 2675** Added "-S" and "--skeleton".
cd5bd6ac 2676
e4ab1254 2677** "%raw", "-r", "--raw" is disabled.
62ab6972 2678
402b123d 2679** Special characters are escaped when output. This solves the problems
cd5bd6ac
AD
2680 of the #line lines with path names including backslashes.
2681
402b123d 2682** New directives.
e4ab1254
AD
2683 "%yacc", "%fixed_output_files", "%defines", "%no_parser", "%verbose",
2684 "%debug", "%source_extension" and "%header_extension".
f987e9d2 2685
402b123d 2686** @$
f987e9d2 2687 Automatic location tracking.
f87a2205 2688\f
402b123d 2689* Changes in version 1.28, 1999-07-06:
d2e00347 2690
402b123d 2691** Should compile better now with K&R compilers.
d2e00347 2692
402b123d 2693** Added NLS.
d2e00347 2694
402b123d 2695** Fixed a problem with escaping the double quote character.
d2e00347 2696
402b123d 2697** There is now a FAQ.
d2e00347 2698\f
402b123d 2699* Changes in version 1.27:
5c31c3c2 2700
402b123d 2701** The make rule which prevented bison.simple from being created on
5c31c3c2
JT
2702 some systems has been fixed.
2703\f
402b123d 2704* Changes in version 1.26:
4be07551 2705
7e508a2b 2706** Bison now uses Automake.
4be07551 2707
402b123d 2708** New mailing lists: <bug-bison@gnu.org> and <help-bison@gnu.org>.
4be07551 2709
402b123d 2710** Token numbers now start at 257 as previously documented, not 258.
4be07551 2711
402b123d 2712** Bison honors the TMPDIR environment variable.
4be07551 2713
402b123d 2714** A couple of buffer overruns have been fixed.
f51dbca1 2715
402b123d 2716** Problems when closing files should now be reported.
f51dbca1 2717
402b123d 2718** Generated parsers should now work even on operating systems which do
f51dbca1 2719 not provide alloca().
4be07551 2720\f
402b123d 2721* Changes in version 1.25, 1995-10-16:
df8878c5 2722
402b123d 2723** Errors in the input grammar are not fatal; Bison keeps reading
df8878c5 2724the grammar file, and reports all the errors found in it.
8c44d3ec 2725
402b123d 2726** Tokens can now be specified as multiple-character strings: for
df8878c5 2727example, you could use "<=" for a token which looks like <=, instead
7e508a2b 2728of choosing a name like LESSEQ.
df8878c5 2729
402b123d 2730** The %token_table declaration says to write a table of tokens (names
df8878c5
RS
2731and numbers) into the parser file. The yylex function can use this
2732table to recognize multiple-character string tokens, or for other
2733purposes.
2734
402b123d 2735** The %no_lines declaration says not to generate any #line preprocessor
df8878c5
RS
2736directives in the parser file.
2737
402b123d 2738** The %raw declaration says to use internal Bison token numbers, not
df8878c5
RS
2739Yacc-compatible token numbers, when token names are defined as macros.
2740
402b123d 2741** The --no-parser option produces the parser tables without including
df8878c5
RS
2742the parser engine; a project can now use its own parser engine.
2743The actions go into a separate file called NAME.act, in the form of
2744a switch statement body.
2745\f
402b123d 2746* Changes in version 1.23:
6780ca7a 2747
4d019228
DM
2748The user can define YYPARSE_PARAM as the name of an argument to be
2749passed into yyparse. The argument should have type void *. It should
2750actually point to an object. Grammar actions can access the variable
2751by casting it to the proper pointer type.
6780ca7a 2752
6780ca7a 2753Line numbers in output file corrected.
6780ca7a 2754\f
402b123d 2755* Changes in version 1.22:
6780ca7a
DM
2756
2757--help option added.
6780ca7a 2758\f
402b123d 2759* Changes in version 1.20:
6780ca7a
DM
2760
2761Output file does not redefine const for C++.
9f4503d6 2762
76551463
AD
2763-----
2764
3209eb1c 2765Copyright (C) 1995-2015 Free Software Foundation, Inc.
76551463 2766
74553c98 2767This file is part of Bison, the GNU Parser Generator.
76551463 2768
f16b0819 2769This program is free software: you can redistribute it and/or modify
76551463 2770it under the terms of the GNU General Public License as published by
f16b0819
PE
2771the Free Software Foundation, either version 3 of the License, or
2772(at your option) any later version.
76551463 2773
f16b0819 2774This program is distributed in the hope that it will be useful,
76551463
AD
2775but WITHOUT ANY WARRANTY; without even the implied warranty of
2776MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2777GNU General Public License for more details.
2778
2779You should have received a copy of the GNU General Public License
f16b0819 2780along with this program. If not, see <http://www.gnu.org/licenses/>.
7e508a2b
AD
2781
2782 LocalWords: yacc YYBACKUP glr GCC lalr ArrayIndexOutOfBoundsException nullptr
2783 LocalWords: cplusplus liby rpl fprintf mfcalc Wyacc stmt cond expr mk sym lr
2784 LocalWords: IELR ielr Lookahead YYERROR nonassoc LALR's api lookaheads yychar
2785 LocalWords: destructor lookahead YYRHSLOC YYLLOC Rhs ifndef YYFAIL cpp sr rr
2786 LocalWords: preprocessor initializer Wno Wnone Werror FreeBSD prec livelocks
2787 LocalWords: Solaris AIX UX RHEL Tru LHS gcc's Wundef YYENABLE NLS YYLTYPE VCG
2788 LocalWords: yyerror cpp's Wunused yylval yylloc prepend yyparse yylex yypush
2789 LocalWords: Graphviz xml nonterminals midrule destructor's YYSTYPE typedef ly
2790 LocalWords: CHR chr printf stdout namespace preprocessing enum pre include's
2791 LocalWords: YYRECOVERING nonfree destructors YYABORT YYACCEPT params enums de
2792 LocalWords: struct yystype DJGPP lex param Haible NUM alloca YYSTACK NUL goto
2793 LocalWords: YYMAXDEPTH Unescaped UCNs YYLTYPE's yyltype typedefs inline Yaccs
2794 LocalWords: Heriyanto Reenable dprec Hilfinger Eggert MYEOF Folle Menezes EOF
242cc08e 2795 LocalWords: Lackovic define's itemset Groff Gettext malloc NEWS'ed YYDEBUG YY
7e508a2b 2796 LocalWords: namespaces strerror const autoconfiguration Dconst Autoconf's FDL
242cc08e
AD
2797 LocalWords: Automake TMPDIR LESSEQ ylwrap endif yydebug YYTOKEN YYLSP ival hh
2798 LocalWords: extern YYTOKENTYPE TOKENTYPE yytokentype tokentype STYPE lval pdf
dcb366b1 2799 LocalWords: lang yyoutput dvi html ps POSIX lvalp llocp Wother nterm arg init
1282c124
AD
2800 LocalWords: TOK calc yyo fval Wconflicts parsers yystackp yyval yynerrs
2801 LocalWords: Théophile Ranquet Santet fno fnone stype associativity Tolmer
2802 LocalWords: Wprecedence Rassoul Wempty Paolo Bonzini parser's Michiel loc
2803 LocalWords: redeclaration sval fcaret reentrant XSLT xsl Wmaybe yyvsp Tedi
2804 LocalWords: pragmas noreturn untyped Rozenman unexpanded Wojciech Polak
6574576c 2805 LocalWords: Alexandre MERCHANTABILITY yytype
7e508a2b
AD
2806
2807Local Variables:
2808mode: outline
e4ab1254 2809fill-column: 76
7e508a2b 2810End: